A hierarchical experimental equipment knowledge graph construction method for complex devices

Abstract

The application discloses a kind of hierarchical experimental equipment knowledge graph construction methods for complex device, the method includes three major links of prelink, construction link and fusion link, the prelink includes analyzing the different levels of experimental equipment in complex device, from different levels under a variety of data sources extraction experimental equipment related entity, experimental equipment entity relationship, experimental equipment attribute information;The construction link includes the construction experimental equipment ontology model and the single level experimental equipment knowledge graph corresponding to different levels by the triple data extracted;The fusion link includes triple data entity alignment, the triple data after alignment is mapped to experimental equipment ontology model, the data in experimental equipment ontology model is matched to the entity in knowledge graph by entity link.The hierarchical experimental equipment knowledge graph constructed by the method has comprehensiveness, stereoscopic, easy to index, solves the problem that the correlation between each device in the experimental process of complex device is difficult to express.

Description

Technical Field

[0001] This invention belongs to the field of machine learning and relates to a knowledge graph construction method, specifically a hierarchical experimental equipment knowledge graph construction method for complex devices. Background Technology

[0002] As human exploration of space deepens, there is an urgent need for in-depth research on the space environment effects of spacecraft materials, devices, and their functional systems.

[0003] A space environment ground simulation device is a complex system with various experimental equipment. The equipment data files within the device are numerous and mostly in document form, resulting in poor intuitiveness and slow retrieval speeds. Furthermore, the experimental systems within the device are highly coupled, with extremely complex business logic. Each system contains a large number of different types of experimental equipment, characterized by large scale, high cost, complex maintenance, and intricate interrelationships. This presents challenges such as difficulty in quickly obtaining basic equipment information, difficulty in locating the source of failure when a piece of equipment malfunctions, and difficulty in rationally planning and utilizing the equipment within the device to provide full lifecycle management and service support. To achieve lifecycle management and service support for the experimental equipment within a space environment ground simulation device, a hierarchical experimental equipment knowledge graph construction method is needed to solve the problem of the difficulty in expressing the relationships between various devices during the experimental process of a complex device. The hierarchical experimental equipment knowledge graph should be comprehensive, three-dimensional, and easily searchable. That is, the constructed hierarchical experimental equipment knowledge graph should be able to comprehensively encompass the basic information of each device in the complex device, the connection methods between devices, the key indicators of the devices, etc., and be presented in a graph structure to express the complexity and hierarchy of the device connections in a good way. At the same time, the hierarchical experimental equipment knowledge graph should be updated according to different functional requirements to reduce redundant node information and improve the retrieval efficiency of experimental equipment information. Summary of the Invention

[0004] To address the challenge of expressing the relationships between various devices during complex experimental procedures, this invention provides a method for constructing a hierarchical experimental equipment knowledge graph for complex devices. The hierarchical experimental equipment knowledge graph constructed by this method is comprehensive, three-dimensional, and easily accessible. It enables lifecycle management and service support for experimental equipment within a ground-based space environment simulation device, presented in a graph structure. This solves the problem of expressing the relationships between various devices during complex experimental procedures and improves the efficiency of information retrieval for experimental equipment.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A hierarchical knowledge graph construction method for complex experimental devices includes three main stages: a pre-construction stage, a construction stage, and a fusion stage.

[0007] The preliminary steps include analyzing different levels that affect the experimental equipment in the complex device, and extracting relevant entities, relationships between experimental equipment entities, and attribute information of the experimental equipment from multiple data sources at different levels.

[0008] The construction process includes constructing an experimental equipment ontology model and a single-level experimental equipment knowledge graph corresponding to different levels by applying the experimental equipment-related entities, relationships between experimental equipment entities, and experimental equipment attribute information extracted from different levels.

[0009] The fusion process includes aligning and linking entities in the constructed single-level experimental equipment knowledge graph, merging entities at different levels in the same domain according to the correlation between entities, constructing a hierarchical experimental equipment knowledge graph, updating the hierarchical experimental equipment knowledge graph according to usage scenarios and functions, and displaying the hierarchical experimental equipment knowledge graph through Neo4j.

[0010] Compared with the prior art, the present invention has the following advantages:

[0011] The hierarchical experimental equipment knowledge graph constructed in this invention encompasses the basic information of each device in the space environment ground simulation device, the connection methods between devices, and the key indicators of the devices, thus possessing comprehensiveness. Presenting the above experimental equipment information in a graph structure effectively expresses the complexity and hierarchy of the device connections, providing a three-dimensional and intuitive representation. Furthermore, the hierarchical experimental equipment knowledge graph is updated to meet different functional requirements, eliminating redundant and weakly related nodes to improve the retrieval efficiency of experimental equipment information, thus ensuring ease of access. Attached Figure Description

[0012] Figure 1 A flowchart illustrating a method for constructing a hierarchical knowledge graph for complex experimental devices;

[0013] Figure 2 A flowchart for a method to construct a hierarchical experimental equipment knowledge graph for complex devices. Detailed Implementation

[0014] The technical solution of the present invention will be further described below with reference to the accompanying drawings, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention that do not depart from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention.

[0015] This invention provides a method for constructing a hierarchical knowledge graph of experimental equipment for complex devices, such as... Figure 1 As shown, the method comprises three main stages: a pre-construction stage, a construction stage, and a fusion stage. The pre-construction stage involves analyzing different levels affecting experimental equipment in a complex device, extracting relevant entities, relationships between entities, and attribute information from multiple data sources at different levels. The construction stage involves building an experimental equipment ontology model and single-level experimental equipment knowledge graphs corresponding to different levels using the extracted triplet data. The fusion stage involves aligning triplet data entities, mapping the aligned triplet data to the experimental equipment ontology model, and matching data in the experimental equipment ontology model to entities in the knowledge graph through entity links. The construction stage utilizes all information obtained in the pre-construction stage to form independent single-level knowledge graphs. The fusion stage links entities in the single-level knowledge graphs from the construction stage, merging entities at different levels within the same domain according to their interrelationships to construct a multi-level knowledge graph. Figure 2 As shown, the specific steps are as follows:

[0016] Step S1: Analyze the different aspects affecting the experimental equipment in the complex apparatus. The specific steps are as follows:

[0017] Analyze the hierarchical structure of complex devices and the scenarios and functions used in the knowledge graph of experimental equipment for complex devices. Based on the above, determine the number of levels in the hierarchical knowledge graph, where:

[0018] The complex device should include multiple experimental systems, which are highly coupled and have extremely complex business logic.

[0019] The experimental equipment knowledge graph is used in two scenarios: monitoring and operation management. Its functions include, but are not limited to, equipment location query, equipment number query, and equipment usage status query.

[0020] Step S2: Determine the data sources at each level, where:

[0021] The data sources include, but are not limited to, specification manuals, product manuals, asset information, service manuals, etc.

[0022] The data sources can be categorized into structured data, semi-structured data, and unstructured data based on their degree of structuring. Specifically, structured data includes, but is not limited to, equipment data and equipment lists in relational databases; semi-structured data includes, but is not limited to, equipment webpage data and manually labeled equipment data; and structured data includes, but is not limited to, equipment specification manuals and equipment image data.

[0023] Step S3: Extract experimental equipment-related entities from multiple data sources at different levels, including:

[0024] The experimental equipment-related entities include the Chinese name of the experimental equipment within the device, the foreign language name of the experimental equipment, and the ID number of the experimental equipment.

[0025] The ID of the experimental equipment consists of 5 parts: system or subsystem name, subsystem name, equipment name, equipment serial number in the device, and sub-equipment serial number. The order of the parts is: system or subsystem name - subsystem name - equipment name - equipment serial number in the device - sub-equipment serial number. The system or subsystem / subsystem / equipment name is named according to the first letter of its Chinese pinyin, and the parts are connected by "-".

[0026] Step S4: Extract the relationships between experimental equipment entities from multiple data sources at different levels, including:

[0027] The relationships between experimental equipment include static and dynamic relationships. The static relationships include, but are not limited to, the size, function, and performance indicators of the experimental equipment. The dynamic relationships include, but are not limited to, the geographical location, working status, and interface usage of the experimental equipment.

[0028] Step S5: Extract experimental equipment attribute information from multiple data sources at different levels, including:

[0029] The experimental equipment relationships and attribute information include, but are not limited to, the geographical location of the experimental equipment, the performance indicators of the experimental equipment, the specifications of the experimental equipment, and the parameters collected by the experimental equipment;

[0030] The experimental equipment-related entities, relationships between experimental equipment entities, and attribute information extracted in steps S6, S3, S4, and S5 constitute triplet data. A knowledge graph ontology model is constructed using this triplet data, wherein:

[0031] The level mentioned includes the system level within the device or the device's business requirements;

[0032] The knowledge graph ontology model includes, but is not limited to, the formulation and connection relationships of rules such as devices, device parameters, connected devices, services, standards, and industry documents, thereby enabling the reuse of individual classes and attributes of a certain ontology.

[0033] Step S7: Construct a single-level knowledge graph of experimental equipment at different levels using triplet data, where:

[0034] The knowledge graph is constructed in two ways: top-down and bottom-up.

[0035] Step S8: Perform entity alignment on the triplet data. The specific steps are as follows:

[0036] The extracted triplet data is aligned to entities, and duplicate and unnecessary nodes are removed—that is, nodes that can be safely removed without disrupting the connectivity of the knowledge graph. This makes the constructed knowledge graph more concise and clear.

[0037] The entity alignment includes entity disambiguation and coreference resolution. Entity disambiguation refers to the fact that one entity reference can correspond to multiple real-world entities; coreference resolution refers to different representations of the same entity.

[0038] Step S9: Map the data stored in the unified data model to the aligned triples;

[0039] Step S10: Merge the various single-level knowledge graphs through entity links. The specific steps are as follows:

[0040] The standard triplet data in each single-level knowledge graph are linked into entities. Based on the relationships between entities, entities at different levels within the same domain are merged together to form a multi-level knowledge graph with a three-dimensional structure and comprehensive content.

[0041] The entity linking includes entity disambiguation, coreference resolution, and associative linking of entities that are related between levels;

[0042] Step S11: Implement a hierarchical knowledge graph display of experimental equipment using Neo4j. The specific steps are as follows:

[0043] This paper uses Python to represent standard data triples and establishes a connection with the Neo4j graph database via API to realize a hierarchical knowledge graph display of experimental equipment.

[0044] The hierarchical knowledge graph display of the experimental equipment is implemented through the graph database Neo4j to adapt to the ever-changing business needs and large-scale data growth in complex devices, and has strong adaptability while maintaining high query performance.

[0045] Example:

[0046] This embodiment provides a method for constructing a hierarchical experimental equipment knowledge graph for complex devices, the method including the following steps:

[0047] Step S1: Analyze the different aspects affecting the experimental equipment in the complex apparatus. The specific steps are as follows:

[0048] Analyze the hierarchical structure of the space environment ground simulation device and the scenarios and functions used in the knowledge graph of the experimental equipment of the space environment ground simulation device. Based on the above, determine the number of levels of the hierarchical knowledge graph.

[0049] In this embodiment, the hierarchical structure of the space environment surface simulation device includes a three-level structure of system / subsystem-subsystem-unit, which can more intuitively reflect the structural hierarchy of the experimental system / supporting system or subsystem. Among them, the system or subsystem includes, but is not limited to, numerical simulation and central monitoring system (hereinafter referred to as "numerical simulation central control system"), comprehensive environmental simulation subsystem (hereinafter referred to as "comprehensive subsystem"), space life science subsystem (hereinafter referred to as "life subsystem"), device ion irradiation subsystem (hereinafter referred to as "device subsystem"), microscopic mechanism analysis subsystem (hereinafter referred to as "microscopic subsystem"), ion accelerator subsystem, space magnetic environment simulation and research system (hereinafter referred to as "magnetic system"), space plasma environment simulation and research system (hereinafter referred to as "plasma system"), low-energy accelerator control system, cooling water control system, building monitoring system, air conditioning control system, radiation protection system, building information model (BIM), power monitoring system, distributed air conditioning monitoring system, video system, access control system, and telephone system.

[0050] In this embodiment, the knowledge graph of the space environment surface simulation device experimental equipment is used in two scenarios: monitoring and operation management. The functions include, but are not limited to, equipment location query, equipment number query, and equipment usage status query.

[0051] Step S2: Determine the original data source for each level in step S1.

[0052] In this embodiment, the data source can be divided into structured data, semi-structured data, and unstructured data according to the degree of structuring. The structured data includes, but is not limited to, equipment data and equipment lists in relational databases; the semi-structured data includes, but is not limited to, equipment web page data and manually labeled equipment data; and the structured data includes, but is not limited to, equipment specification manuals and equipment image data.

[0053] Step S3: Extract relevant entities of experimental equipment from multiple data sources at different levels.

[0054] In this embodiment, experimental equipment entities are extracted from the data source determined in step S2.

[0055] In this embodiment, data integration is used to extract experimental equipment entities for structured data; the BERT-BiLSTM-CRF model is used to extract experimental equipment entities for semi-structured and structured data.

[0056] In this embodiment, the experimental equipment related entities include, but are not limited to, the Chinese name of the experimental equipment, the English name of the experimental equipment, and the experimental equipment ID. Among them, the Chinese name of the experimental equipment includes, but is not limited to, switch, vacuum meter, molecular pump, etc.; the English name of the experimental equipment corresponds to Switch, Vacuum Meter, Pump, etc.; the experimental equipment ID corresponds to SFZK-SYJK-JHJ-9-1, ZH-ZHFZ-ZKJ-284-1, QJ-GNLZFZZD-FZB-360-1.

[0057] In this embodiment, the experimental equipment ID can be easily divided into various levels by the computer. It consists of 5 parts, namely, system or subsystem name, subsystem name, equipment name, equipment serial number in the device, and sub-equipment serial number. The order of each part is: system or subsystem name - subsystem name - equipment name - equipment serial number in the device - sub-equipment serial number. Among them, the system or subsystem / subsystem / equipment name is named according to the first letter of its Chinese pinyin, and each part is connected by "-".

[0058] Step S4: Extract the relationships between experimental equipment entities from multiple data sources at different levels.

[0059] In this embodiment, in order to obtain the triples required to construct the experimental equipment knowledge graph, experimental equipment relationships are extracted from the data source determined in step S2.

[0060] In this embodiment, the experimental equipment relationships include static and dynamic relationships, wherein: the static relationships include, but are not limited to, experimental equipment size, experimental equipment function, experimental equipment performance indicators, etc.; the dynamic relationships include, but are not limited to, experimental equipment geographical location, experimental equipment working status, experimental equipment interface usage, etc.

[0061] Step S5: Extract experimental equipment attribute information from multiple data sources at different levels.

[0062] In this embodiment, experimental equipment attribute information corresponding to entity relationships is extracted from the data source determined in step S2.

[0063] In this embodiment, the experimental equipment attribute information of the entity correspondence includes static information and dynamic information, wherein: the static information includes, but is not limited to, a length-to-height ratio of 16:9, 2U rack-mount, 534 mm * 271 mm * 223 mm, and a switching capacity of 688Gbps; the dynamic information includes, but is not limited to, being located in the computer room on the first floor of the comprehensive building, operating at full load, operating at no load, and using one KVM port.

[0064] S6. Construct a knowledge graph ontology model using the extracted triplet data.

[0065] In this embodiment, the information extracted through steps S3, S4, and S5 is used to form a data triplet, and the upper-level pattern layer framework, i.e., the knowledge graph ontology model, is constructed based on the triplet information.

[0066] In this embodiment, the knowledge graph ontology model includes, but is not limited to, the formulation and connection relationships of rules such as devices, device parameters, connected devices, services, standards, and industry documents, thereby enabling the reuse of individual classes and attributes of a certain ontology.

[0067] Step S7: Construct a single-level knowledge graph based on different levels using the extracted triplet data.

[0068] In this embodiment, the information extracted through steps S3, S4, and S5 is used to form a data triplet, and a lower-level data layer framework is constructed based on the triplet information, namely, a single-level knowledge graph at different levels, which represents knowledge in the form of a graph.

[0069] Step S8: Perform entity alignment on the extracted triplet data.

[0070] In this embodiment, entity alignment is performed on the extracted triplet data, and duplicate and unnecessary nodes are deleted, that is, those nodes that can be safely deleted without breaking the connectivity of the knowledge graph, making the constructed knowledge graph more concise and clear.

[0071] In this embodiment, entity alignment includes entity disambiguation and coreference resolution. Entity disambiguation refers to the fact that one entity reference can correspond to multiple real-world entities. In this embodiment, the device client can correspond to both a PC (computer) and a PLC (programmable logic controller). Coreference resolution refers to different representations of the same entity. In this embodiment, the device vacuum gauge can also be represented as a vacuum meter. This embodiment achieves entity alignment by calculating the similarity after weight constraints and attribute embedding.

[0072] Step S9: Map the data stored in the unified data model to the aligned triples.

[0073] In this embodiment, the data stored in the unified data model is mapped to the triples aligned in step S8 to form a standard data representation.

[0074] In this embodiment, the unified data model is global, fundamental, and basic. That is, the established unified data model is universal and applicable to various complex systems. It can solve the problem of data interconnection and interoperability. At the same time, it can be reconstructed based on the established unified data model to meet different functional requirements.

[0075] In this embodiment, when the unified data model uses TCP at the transport layer and IPv6 at the network layer, the data transmission efficiency should be around 80%, thereby ensuring the efficiency of knowledge graph construction.

[0076] Step S10: Integrate the knowledge graphs at each single level through entity links.

[0077] In this embodiment, the standard triplet data in each single-level knowledge graph are linked as entities. Based on the correlation between entities, entities at different levels in the same domain are merged together to form a multi-level knowledge graph with a three-dimensional structure and comprehensive content.

[0078] Step S11: Implement a hierarchical knowledge graph display of experimental equipment using Neo4j.

[0079] In this embodiment, the standard data triples are represented using the Python language, and a connection is established with the Neo4j graph database via API to realize the hierarchical knowledge graph display of experimental equipment.

Claims

1.A method for constructing a hierarchical experimental equipment knowledge graph for a complex device, characterized in that The method comprises the following steps: Step S1, analyze different levels of experimental equipment in complex devices, the specific steps are as follows: analyze the hierarchical structure of the complex device and the scene and function used by the experimental equipment knowledge graph of the complex device, determine the number of hierarchical knowledge graph levels, the scene used by the experimental equipment knowledge graph is divided into two scenes: monitoring and operation management, and the function includes equipment location query, equipment number query, and equipment usage state query; Step S2, determine the data source under each level; Step S3, extract experimental equipment related entities from multiple data sources under different levels, including Chinese name of experimental equipment in the device, English name of experimental equipment, and ID number of experimental equipment; the ID of the experimental equipment is composed of five parts: system or subsystem name, subsystem name, device name, device serial number in the device, and sub-device serial number, and the arrangement order of each part is: system or subsystem name-subsystem name-device name-device serial number in the device-sub-device serial number, wherein: the system or subsystem / device name is named by the capital letters of its Chinese pinyin, and each part is connected by "-". Step S4, extract experimental equipment entity relationship from multiple data sources under different levels; Step S5, extract experimental equipment attribute information from multiple data sources under different levels; Step S6, the experimental equipment related entities, experimental equipment entity relationship, and experimental equipment attribute information extracted from steps S3, S4, and S5 constitute triple data, and a knowledge graph ontology model is constructed through the triple data; Step S7, construct single-level experimental equipment knowledge graphs corresponding to different levels through triple data; Step S8, align entities of triple data; Step S9, map data stored in a unified data model to aligned triple data; Step S10, fuse single-level knowledge graphs through entity linking; Step S11, realize hierarchical experimental equipment knowledge graph display through Neo4j. 2.The method of claim 1, wherein In step S2, the data sources are divided into structured data, semi-structured data, and unstructured data according to the degree of structuring, wherein: the structured data includes device data in a relational database and a device list; the semi-structured data includes device web page data and device manually annotated data; and the structured data includes device specification manuals and device image data. 3.The method of claim 1, wherein In step S7, the construction method of the knowledge graph includes top-down and bottom-up construction methods. 4.The method of claim 1, wherein The specific steps of step S8 are as follows: aligning entities of extracted triple data, deleting duplicate and unnecessary nodes, wherein: the entity alignment includes entity disambiguation and coreference resolution. 5.The complex device-oriented hierarchical experimental equipment knowledge graph construction method of claim 1, characterized in that The specific steps of step S10 are as follows: linking entities of standard triple data in each single-level knowledge graph, merging entities of different levels existing in the same field according to the relevance between entities, and fusing into a multi-level knowledge graph with structured three-dimensional and comprehensive content, wherein: the entity linking includes entity disambiguation, coreference resolution, and relevance connection between entities of different levels. 6.The complex device-oriented hierarchical experimental equipment knowledge graph construction method of claim 1, characterized in that The specific steps of the step S11 are as follows: using a python language to represent standard data triples, and connecting with a Neo4j graph database through an API to realize hierarchical experimental equipment knowledge graph display.

Images