Standard association method and system based on knowledge graph

Abstract

The application relates to the technical field of knowledge graph, in particular to a standard association method and system based on a knowledge graph; the method comprises the following steps: uniformly accessing multiple-source standard documents and constructing a knowledge evolution unit in a unified expression format; performing multi-layer engineering semantic analysis on the knowledge unit and extracting evolution characteristics; constructing a knowledge evolution fingerprint and establishing a continuity calibration mechanism with the text separated from the knowledge identity; identifying evolution events and forming a knowledge evolution event chain; automatically establishing a standard association relationship according to the transmission law of the event chain among standards; converting the knowledge evolution event chain into a knowledge evolution potential energy and constructing a potential energy field covering the standard association network; and dynamically adjusting and continuously optimizing the standard association according to the potential energy migration, gradient change and global convergence state. The application realizes the stable expression of the standard knowledge graph across versions, the automatic generation and explanation of the association relationship based on the complete evolution process, and the continuous and autonomous evolution of the graph with the standard revision.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and more specifically to a standard association method and system based on knowledge graphs. Background Technology

[0002] As standardization efforts across industries deepen and the number of standard documents continues to grow, revisions and updates become increasingly frequent. The need to maintain knowledge connections and consistency between different standards and between different versions of the same standard is becoming increasingly prominent. Standard documents, as technical specifications and engineering bases, possess characteristics such as rigorous logic, standardized terminology, and clear structure. However, they also face complexities such as adjustments in expression methods, additions, deletions, and merging of clauses, and cross-standard cross-references during version updates. This poses significant challenges to the unified expression, long-term tracking, and construction of connections for standard knowledge. Knowledge graph technology has made significant progress in the fields of general knowledge organization and intelligent retrieval in recent years, providing an effective means to construct semantic connections between standard knowledge.

[0003] Chinese invention patent application CN115640406A discloses a method for multi-source heterogeneous big data analysis and processing and knowledge graph construction, which includes the following steps: First, structured data, semi-structured data and unstructured data are obtained by means of publicly known information data and network retrieval data, respectively, to form a preliminary knowledge graph data hybrid library. Then, massive data sources are obtained through a wide range, multiple fields and various channels. Next, knowledge extraction is performed on the various massive databases obtained for pre-knowledge extraction processing. Then, a more accurate knowledge graph database is obtained by constructing a domain ontology library and a global ontology library in sequence. Finally, after entity alignment and entity linking rule processing, it is provided to users for operation and management in the form of an application platform.

[0004] However, the unique characteristics of version evolution and the continuity of technical knowledge in the field of standard documents require that knowledge graphs not only have static relational expression capabilities, but also be able to reflect the changes in knowledge as standards are revised. How to maintain the stability of knowledge identity amidst the changes in the expression of standard documents, how to organize scattered revision records into an orderly evolutionary path, and on this basis to realize the automatic generation and continuous maintenance of the relationship between standards are important directions worthy of in-depth exploration in the current field of standardization knowledge management. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a standard association method and system based on knowledge graphs.

[0006] The technical solution of this invention: a standard association method based on knowledge graphs, comprising the following specific implementation steps: S1. Perform unified access, structure restoration, complete knowledge segmentation, boundary correction, and source identification management on multi-source standard documents to construct knowledge evolution units with a unified expression format; S2. Based on the knowledge evolution unit and the knowledge evolution source identifier, perform multi-layer engineering semantic analysis on the knowledge unit and extract evolution features to construct a knowledge evolution fingerprint with cross-standard and cross-version continuity. Correct local semantic changes through a knowledge identity continuity calibration mechanism, establish a knowledge evolution fingerprint database and a unified mapping relationship, and realize the separation of knowledge identity and text expression. S3. Based on the knowledge evolution fingerprint and knowledge identity mapping relationship, identify the continuous evolution process of knowledge in different standards and versions, generate a knowledge evolution event chain with time sequence and causal relationship, and automatically generate standard association relationship according to the continuous transmission relationship between the knowledge evolution event chain and the standard knowledge graph that can be dynamically updated. S4. Convert the knowledge evolution event chain into a corresponding knowledge evolution potential energy and maintain a unique binding relationship with the knowledge evolution event chain. Construct a knowledge evolution potential energy field covering the entire standard association network using the standard association network as the propagation carrier. Dynamically adjust and continuously optimize the standard association relationship based on the changes in the comprehensive knowledge evolution potential energy distribution, the local potential field balance state, and the global stable state, so as to realize the autonomous evolution of the knowledge graph with the standard revision process.

[0007] Preferably, knowledge evolution units with a unified expression format are constructed, specifically including: A standard logical skeleton tree is constructed based on the chapter hierarchy, clause number, appendix and figure reference relationship in the standard document, and the structural integrity of logical nodes is calculated according to the continuity of the table of contents index, the continuity of chapter number and the continuity of page number. When the structural integrity is lower than the set requirements, the missing logical nodes are automatically restored. Based on logical nodes, five types of knowledge elements are identified: constraint objects, constraint behaviors, technical indicators, limiting conditions, and applicable scenarios. The knowledge integrity coefficient is calculated based on the completeness of the five types of knowledge elements. When the knowledge integrity coefficient is lower than the set requirement, adjacent clauses are automatically expanded along the logical skeleton tree until a complete knowledge unit is formed. Establish a standardized context continuity analysis mechanism to determine the knowledge context continuity state based on the continuity of reference relationships, logical hierarchy, and technical objects, and perform boundary fusion or boundary preservation processing on adjacent knowledge content according to the context continuity state. A unique knowledge evolution source identifier is generated for each knowledge evolution unit. A mapping relationship is established between standard documents, logical nodes, knowledge evolution units, and knowledge evolution source identifiers. Standard metadata is stored in a relational database, and knowledge evolution units and their relationships are stored in a graph database.

[0008] Preferably, multi-layer engineering semantic parsing is performed on knowledge units and evolutionary features are extracted, specifically including: Using knowledge evolution units as the smallest analysis object, the BERT-BiLSTM-CRF model is used to identify technical objects, constraint behaviors, indicator boundaries, limiting conditions and applicable scenarios. The technical object layer, constraint behavior layer, indicator boundary layer, limiting condition layer and applicable scenario layer are established according to engineering semantics. The unified mapping of synonyms is completed by combining the standard terminology specification mapping library. Features of technical objects, constraint behaviors, indicator boundaries, applicable scenarios, and normative contexts are extracted respectively. Based on the consistency of various features, a knowledge stability evaluation result is established. When the set stability conditions are met, a unique knowledge evolution fingerprint is generated according to the unified coding rules. Using the knowledge evolution source identifier as the anchor point, the consistency of knowledge evolution fingerprints in different standard versions is compared. When the knowledge consistency meets the set requirements, the original knowledge identity is maintained and only the local features that have changed are updated. When the knowledge consistency does not meet the set requirements, a new knowledge identity is generated and an evolutionary mapping relationship between the previous and current knowledge identities is established. Establish a knowledge evolution fingerprint database, and use the knowledge evolution fingerprint as the core identity node to establish a unified mapping relationship with the knowledge evolution unit, knowledge evolution source identifier and standard version, while recording knowledge life cycle information.

[0009] Preferably, the process of identifying the continuous evolution of knowledge across different standards and versions and generating a knowledge evolution event chain specifically includes: Based on the knowledge evolution source identifier, knowledge evolution fingerprint, and standard version information, knowledge identities with continuous source relationships are organized in chronological order to form a candidate evolution sequence; The continuous changes of technical objects, constrained behaviors, indicator boundaries, applicable scenarios and normative contexts in the candidate evolution sequence are analyzed, and the evaluation results of knowledge evolution continuity are established based on the degree of continuous change. When the continuous evolution condition is met, they are connected to form a candidate chain of knowledge evolution events. Identify the knowledge inheritance events, knowledge reinforcement events, knowledge weakening events, knowledge fusion events, knowledge fission events, and knowledge termination events corresponding to each connection in the candidate chain. Establish event type, source standard, target standard, change dimension, change magnitude, occurrence time, and duration attributes for each event, and form a knowledge evolution event chain with causal relationship in chronological order. Based on the number of knowledge evolution events that form a continuous transmission relationship between different standards, a standard association relationship is established, and a multi-layer association network consisting of a knowledge evolution fingerprint layer, a knowledge evolution event layer, a knowledge evolution event chain layer, and a standard association layer is constructed. When a new standard is added or a standard is revised, only the affected local knowledge evolution event chain is updated, and the corresponding complete evolution trajectory is saved simultaneously.

[0010] Preferably, the types of knowledge evolution events to be identified specifically include: When the technical object remains consistent and the technical indicators become more stringent, it is identified as a knowledge reinforcement event; when the technical object remains consistent and the technical indicators become more lenient, it is identified as a knowledge weakening event; when multiple knowledge evolution fingerprints jointly form a new knowledge evolution fingerprint, it is identified as a knowledge fusion event; when a knowledge evolution fingerprint evolves into multiple new knowledge evolution fingerprints, it is identified as a knowledge fission event; when a knowledge evolution fingerprint is no longer inherited by subsequent standards, it is identified as a knowledge termination event; when the knowledge content remains consistent, and only the standard number, standard version, or publication information changes, it is identified as a knowledge inheritance event.

[0011] Preferably, the knowledge evolution event chain is converted into knowledge evolution potential energy and a knowledge evolution potential energy field is constructed, specifically including: Read the event type, evolution span, duration and standard information involved in the knowledge evolution event chain, determine the corresponding initial knowledge evolution potential energy according to the evolution contribution, occurrence time and revision rules of the standard system to which each knowledge evolution event belongs, keep the initial knowledge evolution potential energy uniquely bound to the corresponding knowledge evolution event chain, and establish the mapping relationship between the knowledge evolution potential energy and the standard node. Using the standard association network as the propagation carrier, the knowledge evolution potential energy is propagated layer by layer along the association path formed by the knowledge evolution event chain. The propagation potential energy is attenuated step by step according to the propagation distance and association level. When multiple propagation directions converge to the same standard node, they are accumulated to form the comprehensive knowledge evolution potential energy of the corresponding standard node. Based on the spatial distribution of the comprehensive knowledge evolution potential energy of standard nodes and the local potential field equilibrium state, the standard association relationship is dynamically migrated, the association level is adjusted, and the association strength is optimized. When the standard nodes maintain a strong knowledge evolution trend, the corresponding association relationship is strengthened. When the knowledge evolution trend weakens, the association level is reduced. The basis for association adjustment and the complete evolution trajectory are saved simultaneously. Continuously monitor the changes in the comprehensive knowledge evolution potential energy of each standard node, establish a global stable state evaluation result based on the overall potential energy change state during continuous propagation, stop the current round of potential energy propagation when the preset stability condition is reached, and output the current standard association network, knowledge evolution potential energy field and knowledge graph evolution snapshot; otherwise, continue to execute knowledge evolution potential energy propagation and association relationship adjustment.

[0012] Preferably, based on the spatial distribution of the evolution potential energy of the standard node and the local potential field balance state, the standard association relationship is dynamically migrated, the association level is adjusted, and the association strength is optimized. Specifically, this includes: taking each standard node as the center, calculating its own comprehensive potential energy and the comprehensive potential energy of adjacent nodes, calculating the difference between its own potential energy and the average potential energy of adjacent nodes and normalizing it to obtain the local potential field balance coefficient, initiating local association structure adjustment when the balance coefficient is continuously lower than a set threshold, and keeping the current association structure unchanged when the balance coefficient reaches a stable range. During the adjustment process, only the local area affected by the potential energy field is optimized.

[0013] Preferably, in the calculation of the initial knowledge evolution potential energy, the evolution amplitude of each event comprehensively reflects the degree of adjustment of technical indicators, the degree of change of applicable scope, the degree of change of constraint method, and the degree of change of knowledge structure; the evolution decay coefficient is preset according to the average revision cycle of the standard system to which it belongs.

[0014] Preferably, when dynamically adjusting and continuously optimizing the standard association, the evolution record of each round of knowledge graph and the trajectory of potential field changes are saved simultaneously, so that any standard association can be traced back to its formation basis, adjustment reasons and evolution process.

[0015] The technical solution of the present invention is a standard association system based on knowledge graphs, characterized in that it includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the aforementioned standard association method based on knowledge graphs.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs a standard association method and system based on knowledge graphs. By constructing knowledge evolution units and generating knowledge evolution fingerprints that are separate from textual expressions, the same technical content can maintain a consistent knowledge identity in different standard versions or document formats. This solves the problems of semantic breaks and redundant knowledge definitions caused by standard revisions, significantly enhancing the continuity and stability of knowledge expression. Standard association relationships are established based on knowledge evolution event chains rather than static text similarity. By identifying evolution events such as knowledge inheritance, reinforcement, weakening, fusion, and fission, the associations between standards have clear causal and temporal basis and complete formation trajectories. This not only improves the accuracy of association results but also makes each association traceable and explainable. The introduction of a knowledge evolution potential energy field mechanism enables the standard knowledge graph to automatically adjust local associations and optimize levels based on potential energy propagation and gradient changes. This ensures the timeliness and stability of the graph while significantly reducing maintenance costs. Attached Figure Description

[0017] Figure 1This is a flowchart of a standard association method based on knowledge graphs proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes a standard association method based on knowledge graphs, which includes the following specific implementation steps: S1. By uniformly accessing, restoring the structure, completely segmenting the knowledge, correcting the boundaries, and managing the source identifiers of multi-source standard documents, a Knowledge Evolution Unit (KEU) with a unified expression format is constructed. Based on the logic of standard specifications, a unified knowledge carrier that is traceable and can be associated across versions is formed. S2. Based on the knowledge evolution units and knowledge evolution source identifiers generated in step S1, perform multi-layer engineering semantic analysis on the knowledge units, extract evolution features that can reflect the essential characteristics of knowledge, and construct knowledge evolution fingerprints with cross-standard and cross-version continuity. On this basis, correct local semantic changes through a knowledge identity continuity calibration mechanism, and finally establish a knowledge evolution fingerprint database and a unified mapping relationship to achieve the separation of knowledge identity from text expression. S3. Based on the knowledge evolution fingerprint and knowledge identity mapping relationship generated in step S2, construct a knowledge evolution event chain. By identifying the continuous evolution process of knowledge in different standards and versions, generate a knowledge evolution event chain with time sequence and causal relationship. Based on the transmission law of the event chain between standards, automatically establish standard association relationship, and finally form an event chain driven standard knowledge graph that can be dynamically updated, traced throughout the process, and explained. S4. Based on the knowledge evolution event chain and standard association network generated in step S3, the knowledge evolution event chain is converted into knowledge evolution potential energy with dynamic propagation capability. A knowledge evolution potential energy field covering the entire standard association network is constructed. Based on potential energy migration, potential energy gradient change and overall potential energy convergence state, the standard association relationship is dynamically adjusted and continuously optimized to realize the knowledge graph's autonomous evolution, autonomous stability and full-process traceability as the standard is continuously revised, forming a standard knowledge graph with dynamic adaptability and continuous updating capability.

[0019] In a specific embodiment, step S1 is used to uniformly convert standard documents from different sources, with different structures, and different versions into knowledge evolution units with a consistent expression method. This avoids inconsistencies in knowledge expression caused by differences in standard compilation specifications, format differences, and version updates, and provides a unified data foundation for subsequent knowledge evolution fingerprint construction and standard association analysis. The specific implementation steps include: S11. Unify the access of multiple source documents, including national standards, industry standards, local standards, and international standards, and construct a standard logical skeleton tree based on the relationships between chapters, clauses, appendices, and diagrams to achieve a unified specification structure for standards from different sources. The specific implementation process is as follows: Establish a unified access mechanism for multi-source standard documents, classify and receive standard documents from multiple sources such as national standards, industry standards, local standards, group standards, international standards, and enterprise standards, and automatically identify document types based on document format; For digital native standard documents such as XML, Word, and HTML, the embedded tag information, chapter numbers, and table of contents are directly parsed; for PDF scans, historical photocopies of standards, and image format standards, OCR character recognition is performed first, and after character recognition is completed, page object information such as text blocks, title areas, number areas, table areas, formula areas, figure areas, and footer areas are extracted simultaneously. After extracting the page objects, this embodiment restores the hierarchical relationship between chapters, clauses, appendices, charts and notes according to the writing rules of the standard document itself, and constructs a standard logical skeleton tree. Each node in the standard logical skeleton tree corresponds to a unique logical unit in the standard. The nodes are connected through the chapter hierarchy, so that the entire standard is restored to a tree structure consistent with the original writing specifications. At the same time, the chapter numbering format, clause numbering format, and reference numbering format in different source standards are uniformly converted. For example, different forms of "Chapter 1", "Chapter 1", and "1." are uniformly mapped to the same level of numbering. To evaluate the quality of logical structure recovery, the structural integrity is calculated for each logical node: ; In the formula, This represents the structural integrity of the i-th logical node; This indicates the number of structural elements successfully recovered by the current node, including titles, numbers, body text, tables, formulas, notes, and reference markers, derived from OCR recognition results and digital document parsing results; This indicates the number of structural elements that this type of node should theoretically include, based on the standard template library. This library is derived from the standard template library pre-established by the system, with different templates corresponding to different types of standards. When the structural integrity is lower than the set threshold, it indicates that the current node is missing or has an identification error. At this time, the system will automatically compensate and restore the missing node based on the directory index, the continuity of adjacent chapter numbers and the continuity of page numbers to ensure the continuity and integrity of the entire standard logical structure. S12. Based on the standard logical skeleton tree, knowledge evolution units (KEUs) are generated according to the integrity of technical constraints. By identifying elements such as constraint objects, constraint behaviors, technical indicators, limiting conditions, and applicable scenarios, knowledge units are generated only when a complete technical specification is formed. The specific implementation process is as follows: After completing the construction of the standard logical skeleton tree, each logical node is used as an analysis object to further analyze the technical content inside the node. In this embodiment, knowledge evolution units are generated based on the integrity of the technical constraints. Specifically, each section of technical content is identified with five categories of constraint elements: constraint object, constraint behavior, technical indicators, limiting conditions, and applicable scenarios. Among them, the constraint object refers to the equipment, material, or process object regulated by the standard; the constraint behavior refers to the regulated actions such as "should meet", "must not exceed", and "should be adopted"; the technical indicators refer to specific parameters such as size, pressure, temperature, and accuracy; the limiting conditions refer to the preconditions for the constraint to be valid; and the applicable scenarios refer to the scope of application of the standard or the operating conditions. A knowledge evolution unit is generated only when the above five elements can together constitute a complete technical specification; For example, the standard continuously describes "the transmission pressure shall not exceed 1.6 MPa" and "applicable to natural gas transmission systems", while this embodiment identifies that the two belong to the same technical constraint, and therefore they together form a knowledge evolution unit, so as to keep the technical semantics intact; To evaluate the completeness of knowledge units, a knowledge completeness coefficient is introduced: ; In the formula, This represents the knowledge completeness coefficient. This indicates the number of key constraint elements that have been identified in the current knowledge unit, which is obtained from the statistics of knowledge element identification results; This indicates the number of key elements that a complete technical constraint should include, as determined by the standard knowledge template, which is determined based on the knowledge template corresponding to different standard types. If the knowledge completeness coefficient is lower than the preset threshold, it is considered that the current knowledge expression is not complete. It automatically expands the content of adjacent clauses forward or backward along the logical skeleton tree, recombines the knowledge boundary, until a knowledge evolution unit that can independently express the complete technical specification is formed. S13. For continuous descriptions across clauses, tables, and appendices in the standard, establish a specification context continuity analysis mechanism. This mechanism calculates context continuity by comprehensively considering citation relationships, chapter levels, and the continuity of technical objects. Based on the continuity results, knowledge boundaries are either merged or kept independent, thereby ensuring the engineering semantic integrity of knowledge evolution units. The specific implementation process is as follows: After the knowledge evolution unit is initially formed, this embodiment further establishes a standardized context continuity analysis mechanism to dynamically correct the knowledge boundary; Specifically, the normative reference relationships, chapter hierarchy relationships, and technical object continuity relationships between the current knowledge evolution unit and its adjacent knowledge evolution units are analyzed. When there are continuous references, common constraint objects, or continuous limiting conditions between adjacent knowledge units, it indicates that they jointly describe the same technical specification. In this case, the knowledge boundary is automatically adjusted so that the relevant content is reassigned to the same knowledge evolution unit. If the analysis results show that adjacent knowledge units have formed independent technical constraints, the boundary remains unchanged. continuity of normative references Logical hierarchy continuity Continuity of technical objects After normalization, the context continuity is calculated as follows: ; In the formula, This represents the contextual continuity between knowledge unit i and knowledge unit j; This indicates the degree of continuity of normative references, which is obtained from statistics on cross-reference relationships, i.e., it originates from the results of cross-reference analysis within the standard; It represents the degree of logical hierarchy continuity, calculated based on the hierarchical distance between two knowledge units in the logical skeleton tree, and is derived from the standard logical skeleton tree; It indicates the degree of continuity of technical objects, obtained from the consistency analysis of constraint objects, and is derived from the results of knowledge element identification. , and These represent the three continuous weights, satisfying... ; When the continuity exceeds a preset threshold, knowledge boundary fusion is automatically performed; when the continuity is below the threshold, knowledge units remain independent, thereby ensuring that each knowledge evolution unit has complete and stable engineering semantics. S14. Generate a unique knowledge evolution source identifier for each knowledge evolution unit, and establish a mapping relationship between standard documents, logical nodes, knowledge units, and source identifiers. Simultaneously, store standard metadata in a relational database and knowledge units and their relationships in a graph database to achieve unified management of traceable knowledge sources and sustainable cross-version tracking. The specific implementation process is as follows: After completing the construction and boundary correction of the knowledge evolution unit, in order to ensure that the same technical knowledge can be continuously tracked in the subsequent standard revision process, this embodiment further establishes a knowledge evolution source identifier; Specifically, using standard number, version number, chapter path, knowledge unit location, and knowledge integrity characteristics as basic information, a unique source identifier is generated for each knowledge evolution unit. This source identifier remains stable throughout the lifecycle of the same standard. When the standard is revised, even if the knowledge content is strengthened, weakened, or partially modified, the source of historical knowledge can be quickly located based on the source identifier without having to perform full-text matching again. A one-to-one mapping relationship is established between standard documents, logical skeleton nodes, knowledge evolution units, and knowledge evolution source identifiers, and stored in a relational database and a graph database respectively. The relational database is responsible for storing standard metadata, version information, and mapping indexes; the graph database is responsible for storing the connection relationships and source relationships between knowledge evolution units, and the two are associated through a unified source identifier. In addition, to ensure the traceability of knowledge sources, each knowledge evolution unit synchronously records its source information, such as the standard number, version number, chapter path, formation time, original location, and revision status, forming a complete knowledge source chain. When a new standard is added or a standard is revised, the corresponding knowledge unit can be quickly located based on the source identifier and used as a unified input for knowledge evolution analysis.

[0020] In a specific embodiment, step S2 takes the knowledge evolution unit and knowledge evolution source identifier generated in step S1 as input. While maintaining the complete engineering semantics of the knowledge unit, it performs multi-dimensional semantic parsing on each knowledge evolution unit and constructs a knowledge evolution fingerprint that reflects the essential characteristics of the knowledge. This allows knowledge whose expression changes in different versions of the standard but whose technical connotation remains consistent to obtain a unified identity representation, providing a unified data foundation for subsequent knowledge evolution behavior identification and standard association generation. The specific implementation steps include: S21. Using the knowledge evolution unit as the smallest analysis object, the BERT-BiLSTM-CRF model is used to perform overall semantic parsing of the complete technical constraints, identifying elements such as technical objects, constraint behaviors, indicator boundaries, limiting conditions, and applicable scenarios. These elements are then reorganized according to engineering semantics to form a unified semantic structure. Simultaneously, standard terminology mapping is used to eliminate semantic differences caused by different expressions. The specific implementation process is as follows: Using knowledge evolution units as the smallest semantic analysis granularity, we model the complete technical constraints they contain holistically. Each knowledge evolution unit is input into a BERT-BiLSTM-CRF model that has been continuously trained on standard domain corpora. BERT is used to obtain the overall contextual semantic representation of the knowledge evolution unit, avoiding semantic bias caused by relying solely on local word meanings. BiLSTM is used to analyze the dependencies between different constraint elements within the knowledge evolution unit, ensuring that the technical objects, constraint behaviors, and indicator boundaries remain fully associated. CRF constrains the output results holistically according to the inherent normative expression of the standard document, avoiding problems such as misalignment between technical objects and technical indicators or incorrect identification of constraint behaviors. After completing the basic identification, this embodiment reorganizes the knowledge structure according to engineering semantics; specifically, each knowledge evolution unit is divided into five semantic layers: technical object layer, constraint behavior layer, indicator boundary layer, limiting condition layer and applicable scenario layer, and the dependency relationship between each layer is restored according to the standard logical skeleton tree. For example, for the knowledge evolution unit "the transmission pressure shall not exceed 1.6MPa, applicable to natural gas transmission conditions", "transmission pressure" is identified as the technical object, "shall not exceed" as the constraint behavior, "1.6MPa" as the indicator boundary, and "natural gas transmission conditions" as the applicable scenario. The above content is reorganized into a unified engineering semantic structure, instead of storing multiple entity nodes in a scattered manner. Meanwhile, to avoid inconsistencies in semantic parsing results due to different expression methods in different standards, this embodiment further establishes a standard terminology specification mapping library to uniformly map synonyms, abbreviations, historical terms, and international standard terms; for example, "maximum permissible pressure," "maximum working pressure," and "upper limit of working pressure" are uniformly mapped to the same specification terminology; S22. Based on a unified semantic structure, extract features of technical objects, constraint behaviors, indicator boundaries, applicable scenarios, and normative contexts respectively. Evaluate the stability of each type of feature using a knowledge stability coefficient, and generate a unique knowledge evolution fingerprint according to unified coding rules. The specific implementation process is as follows: After completing multi-layer semantic parsing, this embodiment further condenses the knowledge evolution features of the knowledge evolution unit, extracting five types of evolution features: technical object features, constraint behavior features, indicator boundary features, applicable scenario features, and normative context features. These features are then unified and condensed to form a knowledge evolution identity that can maintain stability over a long period of time. Among them, the technical object characteristics are used to reflect the equipment, materials, processes or testing objects that the knowledge applies to; the constraint behavior characteristics are used to reflect the restriction, permission or recommended methods specified by the standard; the indicator boundary characteristics are used to describe the upper and lower limits, allowable deviations and accuracy range of technical parameters; the applicable scenario characteristics are used to describe the environment, working conditions and industry scope to which the knowledge is applicable; and the specification context characteristics are used to reflect the logical position, reference relationship and upstream and downstream dependency relationship of the knowledge in the standard system. To evaluate the contribution of various features to the stability of knowledge identity, a knowledge stability coefficient is introduced: ; In the formula, Represents the knowledge stability coefficient; This represents the consistency evaluation value of the i-th type of evolutionary feature; Represents the corresponding feature weights, satisfying: m represents the total number of evolutionary features; Among them, the consistency evaluation value of evolutionary features is calculated by the degree of matching between the semantic layers of the same knowledge evolution unit; each weight is obtained by statistical analysis based on historical standard revision data, that is, the frequency of a certain type of feature remaining unchanged in each revision of the standard. The more stable it remains, the higher its corresponding weight. Therefore, the weight comes from the historical standard evolution law, rather than being set by human experience. The knowledge stability coefficient is used to determine whether the knowledge identity has a stable expressive ability. When the knowledge stability coefficient reaches the preset requirement, the five types of evolutionary features are combined in a fixed coding order to generate a knowledge evolution fingerprint, and the knowledge evolution fingerprint is used as the unique knowledge identity for all subsequent standard association analyses. S23. Using the knowledge evolution source identifier as a unified anchor point, compare the consistency of knowledge evolution fingerprints in different versions of the standard, calculate the knowledge consistency index, and perform continuity calibration of knowledge identity based on the comparison results. When the essence of knowledge remains consistent, the original knowledge identity is retained, and only the evolutionary features are updated; when knowledge undergoes substantial changes, a new knowledge identity is generated and an evolutionary mapping relationship is established. The specific implementation process is as follows: Because the revision of standards may involve adding new descriptions, adjusting the expression of indicators, or adding limiting conditions, the same technical knowledge often has local changes in different versions. If the knowledge identity is directly regenerated based on the text, it is easy to misjudge the same knowledge as multiple independent knowledge. Therefore, this embodiment establishes a knowledge evolution fingerprint (KEF) continuity calibration mechanism. Specifically, using the knowledge evolution source identifier generated in step S1 as a unified anchor point, knowledge evolution units with the same source path, the same chapter inheritance relationship, or the same continuous version relationship are automatically grouped into a candidate evolution set. The technical object characteristics, constraint behavior characteristics, indicator boundary characteristics, applicable scenario characteristics, and normative context characteristics in each candidate knowledge evolution fingerprint are compared. When only local features change while other features remain the same, the system does not generate a new knowledge identity, but adds evolution records on the basis of the original knowledge evolution identity to keep the knowledge identity continuous. To evaluate the degree of consistency in knowledge identity, a knowledge consistency index is introduced: ; In the formula, Indicates the knowledge consistency index; This represents the number of knowledge evolution features that maintain consistency. This represents the total number of features contained in a knowledge evolution fingerprint; Among them, the number of consistent knowledge evolution features comes from the item-by-item comparison results between candidate knowledge evolution sets; the total number of features is fixed and determined by the knowledge evolution fingerprint structure. When the knowledge consistency index is higher than the preset threshold, the system retains the original knowledge identity and only updates the local evolutionary features that have changed; when the knowledge consistency index is lower than the threshold, it is determined that new technical knowledge has been formed, and at the same time, an evolutionary mapping relationship between the original knowledge identity and the new knowledge identity is established. S24. Establish a knowledge evolution fingerprint database, using the knowledge evolution fingerprint as the core identity node of the knowledge graph, and establish a unified mapping relationship with the knowledge evolution unit, knowledge evolution source identifier, and standard version. Simultaneously, record knowledge lifecycle information to achieve unified management of knowledge identity, standard version, and source information. The specific implementation process is as follows: After completing the knowledge identity continuity calibration, a knowledge evolution fingerprint database is further established. In this embodiment, the knowledge evolution fingerprint is used as the core node in the entire knowledge graph. Each knowledge evolution fingerprint corresponds to a unique knowledge identity and establishes a unified mapping relationship with the knowledge evolution unit, the knowledge evolution source identifier, and the standard version. Specifically, relational databases continue to be responsible for storing standard metadata, version indexes, knowledge evolution source identifiers, and knowledge identity indexes; graph databases add knowledge evolution fingerprint nodes and establish mapping relationships between knowledge evolution units and knowledge evolution fingerprints, knowledge evolution fingerprints and standard versions, and knowledge evolution fingerprints and knowledge sources, so that knowledge identities can exist long-term without being separated from specific texts. Meanwhile, to ensure that subsequent knowledge evolution analysis has a complete basis, lifecycle information such as the first appearance time, the most recent update time, the standard to which it belongs, the chapter to which it belongs, the source of its formation, the revision status, and the number of historical evolutions is recorded synchronously for each knowledge evolution fingerprint. When subsequent steps S3 perform knowledge inheritance, strengthening, weakening, fusion, and fission analysis, the standard text is no longer directly compared, but the changes between knowledge evolution fingerprints are directly compared, thereby significantly reducing the impact of different text expression methods on the knowledge association results.

[0021] In a specific embodiment, step S3 takes the knowledge evolution fingerprint database and knowledge evolution identity mapping relationship generated in step S2 as input. Instead of directly establishing standard associations based on the similarity between knowledge evolution fingerprints, it first identifies the continuous evolution behavior of knowledge evolution fingerprints during the standard revision process and organizes multiple knowledge evolution events with continuous inheritance relationships into a Knowledge Evolution Event Chain (KEEC). Then, based on the transmission rules of the knowledge evolution event chain among multiple standards, standard association relationships are automatically generated, ensuring that standard associations are established on the basis of continuous knowledge evolution rather than static textual expression. This achieves the transformation of the standard knowledge graph from static associations to dynamic evolutionary associations. The specific implementation steps include: S31. Based on the knowledge evolution source identifier, knowledge evolution fingerprint, and standard version information, knowledge identities with continuous source relationships are organized in chronological order to form a candidate evolution sequence. The continuity of evolutionary characteristics, such as technical objects, constrained behaviors, indicator boundaries, applicable scenarios, and specification context, is comprehensively analyzed to gradually expand and form a candidate chain of knowledge evolution events. The specific implementation process is as follows: Using knowledge evolution fingerprints as the basic object, a candidate chain of knowledge evolution events is constructed under the constraint of knowledge evolution source identifiers; Specifically, based on the knowledge evolution source identifier, knowledge evolution fingerprints belonging to the same standard system, the same technical object, or having a continuous version relationship are collected, and a time series is established according to the standard release date, implementation date, and revision time; Taking each knowledge evolution fingerprint as the starting node, we sequentially search for subsequent knowledge evolution fingerprints in the time series and analyze whether the five types of evolution features—technical object, constraint behavior, indicator boundary, applicable scenario, and normative context—remain continuous. When the above features meet the continuous evolution condition, a candidate evolution connection is established between the two knowledge evolution fingerprints, and the connection continues to expand until a continuous evolution relationship can no longer be formed, thus forming a complete knowledge evolution event candidate chain. It should be noted that this embodiment does not establish a connection based solely on two adjacent versions, but allows knowledge evolution events to be continuously transmitted across multiple standard versions. For example, when a piece of technical knowledge goes through multiple stages such as "initial proposal - indicator strengthening - scope of application expansion - final replacement" during the standard revision process, the continuous relationship between each stage can be fully preserved, and the entire evolution chain will not be interrupted because a certain version has not undergone significant changes. To evaluate whether candidate knowledge evolution event chains possess continuous evolutionary characteristics, evolutionary continuity is introduced: ; In the formula, It represents the evolutionary continuity between adjacent knowledge evolution fingerprints in a knowledge evolution event chain; The degree of continuity of the i-th type of evolutionary feature includes the degree of continuity of technical objects, the degree of continuity of constraint behaviors, the degree of continuity of indicator boundaries, the degree of continuity of applicable scenarios, and the degree of continuity of normative context. The weights representing the corresponding evolutionary features are determined based on the revised statistical results according to historical standards. It should be noted that each degree of continuity is derived from the item-by-item comparison results of the knowledge evolution fingerprint generated in step S2; each weight is derived from the stable statistical frequency of different features in the historical standard version database. When the evolutionary continuity reaches a preset threshold, the system considers the knowledge evolution to be continuous and adds the corresponding connection to the knowledge evolution event candidate chain; otherwise, it stops expanding and searches for a new evolutionary starting point. S32. Identify the evolution type of each knowledge connection in the candidate chain of knowledge evolution events, distinguishing different evolution events such as knowledge inheritance, reinforcement, weakening, fusion, fission, and termination. Simultaneously, establish event attributes for each event, including event type, source standard, target standard, change dimension, change magnitude, occurrence time, and duration. Organize these attributes chronologically to form a causal chain of knowledge evolution events. The specific implementation process is as follows: After completing the construction of the knowledge evolution event candidate chain, this embodiment further identifies the event type of each knowledge evolution connection in the candidate chain and constructs complete knowledge evolution event chain attributes; Specifically, the changes in the knowledge evolution fingerprint before and after a knowledge evolution event are analyzed. When the technical object remains the same, but the technical indicators become more stringent, it is identified as a knowledge reinforcement event. When the technical object remains the same, but the technical indicators are relaxed, it is identified as a knowledge weakening event. When multiple knowledge evolution fingerprints evolve together to form a new knowledge evolution fingerprint, it is identified as a knowledge fusion event. When one knowledge evolution fingerprint evolves into multiple new knowledge evolution fingerprints, it is identified as a knowledge fission event. When a knowledge evolution fingerprint is no longer inherited by subsequent standards, it is identified as a knowledge termination event. When the knowledge content remains the same overall, but only the standard number or version changes, it is identified as a knowledge inheritance event. After completing event identification, event attributes are further established for each knowledge evolution event, including but not limited to event type, occurrence time, source standard, target standard, involved technology object, change dimension, change magnitude, duration and evolution stage information; at the same time, all events in the same knowledge evolution event chain are organized into an event chain structure according to the order of occurrence, so that there is not only a time relationship between events, but also a strict causal transmission relationship. For example, a technical requirement may go through four consecutive events in sequence: "knowledge inheritance → knowledge reinforcement → knowledge fusion → knowledge replacement". These events are stored as a whole as a knowledge evolution event chain, rather than being stored as four isolated events. S33. Using knowledge evolution event chains as the basis for standard association generation, by statistically analyzing the transmission of knowledge evolution event chains in different standards, it analyzes whether the event chains form a continuous evolutionary trajectory across multiple standards, and automatically establishes various standard association relationships such as knowledge inheritance association, knowledge reinforcement association, knowledge fusion association, knowledge fission association, and knowledge synergy association based on the credibility of evolution association, thereby realizing the automatic discovery of implicit knowledge associations. The specific implementation process is as follows: After constructing the knowledge evolution event chain, this embodiment further generates standard associations based on the knowledge evolution event chain. Specifically, it counts all knowledge evolution event chains contained in each standard and constructs standard knowledge evolution trajectories based on the amount of knowledge, duration, and event type involved in the knowledge evolution event chain. It analyzes whether there are continuously transmitted knowledge evolution event chains between different standards. When a knowledge evolution event chain in one standard can naturally continue to another standard, it is considered that there is a real knowledge inheritance relationship between the two standards. When multiple standards share the same knowledge evolution event chain but evolve different subsequent events, a knowledge collaboration association is established. When the knowledge evolution event chains in multiple standards eventually converge to form the same knowledge evolution event, a knowledge fusion association is established. To evaluate the reliability of criterion associations, an evolutionary association reliability is introduced: ; In the formula, Indicates the credibility of the standard association; The number of knowledge evolution events that form a continuous transitive relationship between two standards; This represents the total number of knowledge evolution events across both standards. The number of consecutively transmitted events is derived from the results of the knowledge evolution event chain analysis; the total number of events is derived from the corresponding standard knowledge evolution event set. When the credibility of the evolutionary association reaches the set threshold, standard association edges are automatically established, and different association attributes such as knowledge inheritance association, knowledge reinforcement association, knowledge fusion association, knowledge fission association, and knowledge collaboration association are assigned according to the main event types in the knowledge evolution event chain. Since the association is based on a complete chain of knowledge evolution events, even if there is no direct reference relationship between the two standards, a stable association can be formed based on the long-term evolution of knowledge. S34. Construct a multi-layered association network consisting of a knowledge evolution fingerprint layer, a knowledge evolution event layer, a knowledge evolution event chain layer, and a standard association layer. When adding, revising, or abolishing standards, only the newly added knowledge evolution events and affected event chains are partially updated. At the same time, the complete knowledge evolution event chain corresponding to each standard association is saved, realizing dynamic maintenance, full-process traceability, and interpretable basis for standard associations. The specific implementation process is as follows: After generating the standard association, this embodiment further constructs an event chain-driven association network. The entire association network includes a four-level structure: a knowledge evolution fingerprint layer, a knowledge evolution event layer, a knowledge evolution event chain layer, and a standard association layer. Among them, the knowledge evolution fingerprint layer is used to store knowledge identity; the knowledge evolution event layer is used to record single knowledge changes; the knowledge evolution event chain layer is used to describe the continuous transmission process between multiple knowledge events; and the standard association layer is used to store the final association relationship formed between standards. When a new standard is added, a standard is revised, or a standard is abolished, the new knowledge evolution event is re-identified based on the new knowledge evolution fingerprint, and it is further determined whether the new event can be connected to the existing knowledge evolution event chain. If it can be connected, only the corresponding event chain and the association of the affected standard are updated, without recalculating the entire knowledge graph. If it cannot be connected, a new knowledge evolution event chain is automatically generated, and a new standard association is established. At the same time, the knowledge evolution event chain corresponding to each standard association is recorded synchronously, and all knowledge evolution events, standard versions and formation times involved in the event chain are saved, so as to realize the traceability of the entire standard association process. When a user queries any standard association, he / she can not only obtain the association result, but also view the complete knowledge evolution event chain that formed the association, providing a complete basis for standard revision analysis, standard system research and impact analysis.

[0022] In a specific embodiment, step S4 takes the knowledge evolution event chain and standard association network formed in step S3 as input, regards each knowledge evolution event chain as an evolutionary source that continuously releases the influence of knowledge evolution, and continuously analyzes the generation, propagation, migration and balance process of the influence of knowledge evolution in the standard association network by constructing a knowledge evolution potential energy field. This enables the standard association relationship to automatically adjust according to the knowledge evolution trend, realizing the continuous self-evolution of the standard knowledge graph. The specific implementation steps include: S41. Read information such as event type, evolution span, duration, and involved standards from the knowledge evolution event chain. Treat the entire knowledge evolution event chain as a unified analysis object, calculate the knowledge evolution potential energy, and generate the corresponding knowledge evolution potential energy. At the same time, establish the mapping relationship between the potential energy and the knowledge evolution event chain and standard nodes. The specific implementation process is as follows: Read all knowledge evolution events in each knowledge evolution event chain, and count the event type, evolution span, number of standards involved, and continuous evolution cycle according to the order of event occurrence. Based on the overall evolution characteristics of the event chain, allocate initial knowledge evolution potential energy to each event chain. In order to ensure that the initial potential energy can reflect the continuous evolutionary capacity of knowledge, a knowledge evolution potential energy calculation model is introduced: ; In the formula, This represents the initial knowledge evolution potential energy corresponding to the current knowledge evolution event chain; The evolution range of the i-th knowledge evolution event is derived from the event attributes recorded in step S3, and comprehensively reflects the degree of adjustment of technical indicators, the degree of change of applicable scope, the degree of change of constraint method, and the degree of change of knowledge structure. This represents the time interval between the current time and the occurrence time of the i-th knowledge evolution event, which is derived from the time records in the knowledge evolution event chain; The evolution decay coefficient is used to describe the objective law that the influence of knowledge gradually weakens over time. It can be preset according to the average revision cycle of the standard system to which it belongs; n represents the total number of knowledge evolution events contained in the current knowledge evolution event chain. Based on the above calculations, knowledge evolution events that have occurred recently and have undergone significant changes can acquire higher potential energy, while the potential energy of knowledge events that have not continued to evolve over a long period of time will gradually decrease, making the potential energy of knowledge evolution more consistent with the actual law of continuous revision of standards. After completing the potential energy calculation, the knowledge evolution potential energy is bound to the corresponding knowledge evolution event chain and written into the knowledge evolution event chain node in the graph database. At the same time, a mapping relationship between the potential energy and the corresponding standard node is established. S42. Using the standard association network as the propagation carrier, the knowledge evolution potential energy is propagated layer by layer along the standard association path. The potential energy is gradually attenuated according to the propagation distance, and the potential energy formed by the convergence of multiple propagation directions is accumulated and balanced. This enables each standard node to form a comprehensive knowledge evolution potential energy, and finally constructs a knowledge evolution potential energy field that reflects the overall knowledge evolution trend of the standard system. The specific implementation process is as follows: After completing the construction of the knowledge evolution potential energy, the standard correlation network formed in step S3 is used as the propagation carrier to continuously propagate the knowledge evolution potential energy and form a local knowledge evolution potential field in the standard network. Specifically, each knowledge evolution potential energy releases its evolution potential energy starting from its corresponding standard node, and the potential energy propagates step by step to adjacent standards along the association direction determined by the knowledge evolution event chain. During the propagation process, the propagation potential energy is attenuated according to the propagation distance and association type after each layer of standard association. If multiple knowledge evolution potential energy propagates to the same standard node, their respective propagation results are retained and accumulated at that node to form the comprehensive knowledge evolution potential energy of the current node. To describe the propagation law of knowledge evolution potential energy in standard association networks, a propagation potential energy calculation model is introduced: ; In the formula, This represents the remaining potential energy after the knowledge evolution potential energy propagates to the j-th standard node; This represents the cumulative number of knowledge evolution event chain levels traversed between the knowledge evolution potential energy and the j-th standard node; This represents the propagation attenuation coefficient, used to reflect the law that the influence of knowledge gradually weakens as the propagation distance increases; Continuously record all propagation potential energy received by each standard node and form the comprehensive potential energy value of the corresponding node. The higher the comprehensive potential energy, the more significantly the standard is affected by knowledge evolution in the current standard system, and the higher the priority of its subsequent correlation adjustment. S43. Based on the comprehensive knowledge evolution potential energy and potential energy gradient changes between standard nodes, the standard association relationships are dynamically migrated and their levels are adjusted. When the standard nodes maintain a strong knowledge evolution trend, the association is strengthened; when the knowledge evolution trend gradually weakens, the association level is reduced. At the same time, the basis for association adjustment and the knowledge evolution trajectory are completely preserved, realizing the adaptive optimization of the standard association structure. The specific implementation process is as follows: After the knowledge evolution potential field is formed, the local potential field distribution between each standard node is analyzed to find the position where the influence of knowledge evolution reaches a relatively stable state. Specifically, taking each standard node as the center, its own comprehensive potential energy and the comprehensive potential energy of adjacent standard nodes are statistically analyzed, and it is analyzed whether there is a continuous imbalance in the distribution of knowledge evolution potential energy in the local area. When a certain standard node maintains a significantly higher comprehensive potential energy than the surrounding nodes for a long period of time, it indicates that the current standard has become a new aggregation center of knowledge evolution. Further analysis is conducted to see if there are standard associations that have not yet been established but have a basis for continuous knowledge evolution. When a certain standard node maintains a low comprehensive potential energy for a long period of time and the corresponding knowledge evolution event chain has terminated, it is considered that the importance of the association has decreased, and its association level can be reduced. To describe whether the local potential field at a standard node has reached a steady state, a local potential field equilibrium coefficient is introduced: ; In the formula, This represents the local potential field equilibrium coefficient; This represents the comprehensive knowledge evolution potential of the current standard node; This represents the average combined potential energy of all neighboring standard nodes of the current standard node; This represents the potential energy difference between the current node and its surrounding nodes; When the local potential field balance coefficient remains below the set threshold, it indicates that the distribution of knowledge evolution influence in the current region is uneven, and local correlation structure adjustment is initiated; when the local potential field balance coefficient reaches a stable range, the current correlation structure remains unchanged. Throughout the adjustment process, only the local regions affected by the knowledge evolution potential field are optimized, without changing the association structure of other stable regions, thereby reducing the overall update cost of the map. S44. Continuously analyze the changes in the comprehensive knowledge evolution potential energy and association migration of each standard node, and determine whether the knowledge graph has reached a stable evolutionary state based on the graph stability index; when the stability condition is met, output the latest standard association network, knowledge evolution potential energy field, and evolution snapshot; when the stable state has not been reached, continue to execute potential energy propagation and association migration to achieve continuous self-evolution and stable output of the knowledge graph. The specific implementation process is as follows: After completing the local correlation adjustment, the knowledge evolution potential field is further analyzed from the perspective of the entire standard correlation network to determine whether it has reached an overall stable state. Based on this, it is decided whether to complete this round of knowledge graph self-evolution. Specifically, the changes in the comprehensive potential energy of all standard nodes in the current correlation network are statistically analyzed, and the distribution of the knowledge evolution potential field in two adjacent rounds is continuously compared. When the changes in the comprehensive potential energy of most standard nodes tend to stabilize during two consecutive rounds of propagation, and the local potential field balance coefficient remains within the set range, it is considered that the entire knowledge evolution potential field has completed this round of self-organization process. To describe the current stability of the entire knowledge graph, a global stability index is introduced: ; In the formula, This represents the global stability index of a knowledge graph. This represents the comprehensive knowledge evolution potential of the i-th standard node after the k-th round of propagation; The comprehensive knowledge evolution potential of the i-th standard node after the previous round of propagation is represented; N represents the total number of standard nodes in the current standard association network. When the global stability index reaches the preset stability threshold, the current round of knowledge evolution and propagation is stopped, and the current standard association network, knowledge evolution potential field distribution, knowledge evolution event chain and knowledge evolution fingerprint are uniformly written into the graph database and a new knowledge graph version is generated; if the stability threshold is not reached, steps S42 and S43 are re-executed until the entire knowledge graph reaches a stable state. In addition, the evolution records of each round of knowledge graph and the trajectory of potential field changes are preserved, so that any standard association can be traced back to its formation basis, adjustment reasons and evolution process.

[0023] Example 2: The present invention proposes a standard association system based on a knowledge graph, which is used to execute a standard association method based on a knowledge graph proposed in Example 1, and includes: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement a standard association method based on a knowledge graph, as described in Embodiment 1 above.

[0024] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A standard association method based on knowledge graphs, characterized in that, The specific implementation steps include the following: S1. Perform unified access, structure restoration, complete knowledge segmentation, boundary correction, and source identification management on multi-source standard documents to construct knowledge evolution units with a unified expression format; S2. Based on the knowledge evolution unit and the knowledge evolution source identifier, perform multi-layer engineering semantic analysis on the knowledge evolution unit and extract evolution features to construct a knowledge evolution fingerprint with cross-standard and cross-version continuity. Correct local semantic changes through a knowledge identity continuity calibration mechanism, establish a knowledge evolution fingerprint database and a unified mapping relationship, and realize the separation of knowledge identity and text expression. This includes performing multi-layer engineering semantic parsing on knowledge evolution units and extracting evolutionary features, specifically including: Using knowledge evolution units as the smallest analysis object, the BERT-BiLSTM-CRF model is used to identify technical objects, constraint behaviors, indicator boundaries, limiting conditions and applicable scenarios. The technical object layer, constraint behavior layer, indicator boundary layer, limiting condition layer and applicable scenario layer are established according to engineering semantics. The unified mapping of synonyms is completed by combining the standard terminology specification mapping library. Features of technical objects, constraint behaviors, indicator boundaries, applicable scenarios, and normative contexts are extracted respectively. Based on the consistency of various features, a knowledge stability evaluation result is established. When the set stability conditions are met, a unique knowledge evolution fingerprint is generated according to the unified coding rules. Using the knowledge evolution source identifier as the anchor point, the consistency of knowledge evolution fingerprints in different standard versions is compared. When the knowledge consistency meets the set requirements, the original knowledge identity is maintained and only the local features that have changed are updated. When the knowledge consistency does not meet the set requirements, a new knowledge identity is generated and an evolutionary mapping relationship between the previous and current knowledge identities is established. Establish a knowledge evolution fingerprint database, and use the knowledge evolution fingerprint as the core identity node to establish a unified mapping relationship with the knowledge evolution unit, knowledge evolution source identifier and standard version, while recording knowledge life cycle information; S3. Based on the knowledge evolution fingerprint and knowledge identity mapping relationship, identify the continuous evolution process of knowledge in different standards and versions, generate a knowledge evolution event chain with time sequence and causal relationship, and automatically generate standard association relationship according to the continuous transmission relationship between the knowledge evolution event chain and the standard knowledge graph that can be dynamically updated. S4. Convert the knowledge evolution event chain into the corresponding knowledge evolution potential energy and maintain a unique binding relationship with the knowledge evolution event chain. Construct a knowledge evolution potential energy field covering the entire standard association network using the standard association network as the propagation carrier. Dynamically adjust and continuously optimize the standard association relationship based on the changes in the comprehensive knowledge evolution potential energy distribution, the local potential field balance state, and the global stable state, so as to realize the autonomous evolution of the knowledge graph with the standard revision process. Specifically, this involves converting the knowledge evolution event chain into corresponding knowledge evolution potential energy and constructing a knowledge evolution potential energy field, including: Read the event type, evolution span, duration and standard information involved in the knowledge evolution event chain, determine the corresponding initial knowledge evolution potential energy according to the evolution contribution, occurrence time and revision rules of the standard system to which each knowledge evolution event belongs, keep the initial knowledge evolution potential energy uniquely bound to the corresponding knowledge evolution event chain, and establish the mapping relationship between the knowledge evolution potential energy and the standard node. Using the standard association network as the propagation carrier, the knowledge evolution potential energy is propagated layer by layer along the association path formed by the knowledge evolution event chain. The propagation potential energy is attenuated step by step according to the propagation distance and association level. When multiple propagation directions converge to the same standard node, they are accumulated to form the comprehensive knowledge evolution potential energy of the corresponding standard node. Based on the spatial distribution of the comprehensive knowledge evolution potential energy of standard nodes and the local potential field equilibrium state, the standard association relationship is dynamically migrated, the association level is adjusted, and the association strength is optimized. When the standard nodes maintain a strong knowledge evolution trend, the corresponding association relationship is strengthened. When the knowledge evolution trend weakens, the association level is reduced. The basis for association adjustment and the complete evolution trajectory are saved simultaneously. Continuously monitor the changes in the comprehensive knowledge evolution potential energy of each standard node, establish a global stable state evaluation result based on the overall potential energy change state during continuous propagation, stop the current round of potential energy propagation when the preset stability condition is reached, and output the current standard association network, knowledge evolution potential energy field and knowledge graph evolution snapshot; otherwise, continue to execute knowledge evolution potential energy propagation and association relationship adjustment.

2. The standard association method based on knowledge graphs according to claim 1, characterized in that, Constructing knowledge evolution units with a unified expression format, specifically including: A standard logical skeleton tree is constructed based on the chapter hierarchy, clause number, appendix and figure reference relationship in the standard document, and the structural integrity of logical nodes is calculated according to the continuity of the table of contents index, the continuity of chapter number and the continuity of page number. When the structural integrity is lower than the set requirements, the missing logical nodes are automatically restored. Based on logical nodes, five types of knowledge elements are identified: constraint objects, constraint behaviors, technical indicators, limiting conditions, and applicable scenarios. The knowledge integrity coefficient is calculated based on the completeness of the five types of knowledge elements. When the knowledge integrity coefficient is lower than the set requirement, adjacent clauses are automatically expanded along the logical skeleton tree until a complete knowledge evolution unit is formed. Establish a standardized context continuity analysis mechanism to determine the knowledge context continuity state based on the continuity of reference relationships, logical hierarchy, and technical objects, and perform boundary fusion or boundary preservation processing on adjacent knowledge content according to the context continuity state. A unique knowledge evolution source identifier is generated for each knowledge evolution unit. A mapping relationship is established between standard documents, logical nodes, knowledge evolution units, and knowledge evolution source identifiers. Standard metadata is stored in a relational database, and knowledge evolution units and their relationships are stored in a graph database.

3. The standard association method based on knowledge graphs according to claim 2, characterized in that, Identify the continuous evolution of knowledge across different standards and versions and generate a knowledge evolution event chain, specifically including: Based on the knowledge evolution source identifier, knowledge evolution fingerprint, and standard version information, knowledge identities with continuous source relationships are organized in chronological order to form a candidate evolution sequence; The continuous changes of technical objects, constrained behaviors, indicator boundaries, applicable scenarios and normative contexts in the candidate evolution sequence are analyzed, and the evaluation results of knowledge evolution continuity are established based on the degree of continuous change. When the continuous evolution condition is met, they are connected to form a candidate chain of knowledge evolution events. Identify the knowledge inheritance events, knowledge reinforcement events, knowledge weakening events, knowledge fusion events, knowledge fission events, and knowledge termination events corresponding to each connection in the candidate chain. Establish event type, source standard, target standard, change dimension, change magnitude, occurrence time, and duration attributes for each event, and form a knowledge evolution event chain with causal relationship in chronological order. Based on the number of knowledge evolution events that form a continuous transmission relationship between different standards, a standard association relationship is established, and a multi-layer association network consisting of a knowledge evolution fingerprint layer, a knowledge evolution event layer, a knowledge evolution event chain layer, and a standard association layer is constructed. When a new standard is added or a standard is revised, only the affected local knowledge evolution event chain is updated, and the corresponding complete evolution trajectory is saved simultaneously.

4. The standard association method based on knowledge graphs according to claim 3, characterized in that, Identifying the types of knowledge evolution events specifically includes: When the technical object remains consistent and the technical indicators become more stringent, it is identified as a knowledge reinforcement event; when the technical object remains consistent and the technical indicators become more lenient, it is identified as a knowledge weakening event; when multiple knowledge evolution fingerprints jointly form a new knowledge evolution fingerprint, it is identified as a knowledge fusion event; when a knowledge evolution fingerprint evolves into multiple new knowledge evolution fingerprints, it is identified as a knowledge fission event; when a knowledge evolution fingerprint is no longer inherited by subsequent standards, it is identified as a knowledge termination event; when the knowledge content remains consistent, and only the standard number, standard version, or publication information changes, it is identified as a knowledge inheritance event.

5. The standard association method based on knowledge graphs according to claim 1, characterized in that, Based on the spatial distribution of potential energy and the local potential field balance state of the standard node's comprehensive knowledge evolution, the standard association relationship is dynamically migrated, the association level is adjusted, and the association strength is optimized. Specifically, this includes: taking each standard node as the center, calculating its own comprehensive potential energy and the comprehensive potential energy of adjacent nodes, calculating the difference between its own potential energy and the average potential energy of adjacent nodes and normalizing it to obtain the local potential field balance coefficient, initiating local association structure adjustment when the balance coefficient is continuously lower than a set threshold, and maintaining the current association structure unchanged when the balance coefficient reaches a stable range. During the adjustment process, only the local area affected by the potential energy field is optimized.

6. The standard association method based on knowledge graphs according to claim 1, characterized in that, In the calculation of the initial knowledge evolution potential energy, the evolution amplitude of each event comprehensively reflects the degree of adjustment of technical indicators, the degree of change in the scope of application, the degree of change in the constraint method, and the degree of change in the knowledge structure. The evolution decay coefficient is preset based on the average revision cycle of the relevant standard system.

7. The standard association method based on knowledge graphs according to claim 1, characterized in that, When dynamically adjusting and continuously optimizing standard associations, the evolution record of each round of knowledge graph and the trajectory of potential field changes are saved simultaneously, so that any standard association can be traced back to its formation basis, adjustment reasons and evolution process.

8. A standard association system based on knowledge graphs, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement a standard association method based on a knowledge graph as described in any one of claims 1 to 7.

Images