Knowledge graph construction method, device and equipment for operation and maintenance health management

Abstract

The invention provides a knowledge graph construction method, device and equipment for operation and maintenance health management, and the method constructs a corresponding target knowledge graph based on a fault management specification file and a maintenance management specification file. Wherein the corresponding document fields are extracted from the fault management specification file through the preset content ontology so as to determine the entity nodes in the atlas, and the specification files of different manufacturers can determine the corresponding fields based on the content ontology so as to realize quick alignment of the multi-source data. And meanwhile, on the basis of the knowledge graph constructed by the document fields extracted based on the fault management specification file, corresponding maintenance nodes are supplemented, so that the final target operation and maintenance management knowledge graph not only can position a fault mode, but also can directly determine a corresponding maintenance means, thereby providing reliable data support for operation and maintenance management, and improving the reliability of the operation and maintenance management. The processing efficiency and quality of each operation and maintenance management scene facing sudden abnormal events are improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a method, apparatus and equipment for constructing a knowledge graph for operation and maintenance health management. Background Technology

[0002] For dispersed and complex operation and maintenance scenarios, such as airports and high-speed rail stations, timely fault location and handling are crucial. Related technologies typically rely on operations and maintenance personnel analyzing information from relevant operation and maintenance documents. However, manually extracting fragmented knowledge from numerous operation and maintenance documents and compiling corresponding fault diagnosis and repair solutions is inefficient and fails to meet the needs of handling sudden abnormal events in complex operation and maintenance scenarios. Summary of the Invention

[0003] This disclosure provides a method, apparatus, and equipment for constructing a knowledge graph for operation and maintenance health management, so as to at least solve the above-mentioned technical problems existing in the prior art.

[0004] A first aspect of this disclosure provides a method for constructing a knowledge graph for operational health management, the method comprising: Obtain the fault management specification document and the maintenance management specification document; Extract the corresponding document fields from the fault management specification file according to the preset content ontology, and construct the fault management specification table according to the content ontology and the document fields; The fault management specification table is structurally transformed, and the document fields of each row in the fault management specification table are used as corresponding entity nodes to construct a corresponding knowledge subgraph. The entity nodes include at least fault mode nodes. The similarity of entity nodes in each knowledge subgraph is compared, and the knowledge subgraphs are merged according to the comparison results to obtain the corresponding initial knowledge graph. Extract the fault mode phrases and matching maintenance operation phrases from the maintenance management specification document; Identify the target fault mode node in the initial knowledge graph that corresponds to the fault mode segment, and complete the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph.

[0005] In one possible implementation, the step of extracting corresponding document fields from the fault management specification file according to a preset content ontology, and constructing a fault management specification table based on the content ontology and the document fields, includes: The corresponding column name fields in the fault management specification file are determined based on the preset content ontology; Construct a corresponding ontology template based on the content ontology and the column name field; Identify the document field corresponding to the column name field in the fault management specification file, and replace the column name field in the ontology template with the document field to obtain the corresponding fault management specification table.

[0006] In one possible implementation, before determining the document field corresponding to the column name field in the fault management specification file, the method further includes: Obtain the corrected ontology template, and determine the corresponding document fields based on the column name fields in the corrected ontology template.

[0007] In one possible implementation, constructing a corresponding knowledge subgraph by using the document fields of each row in the fault management specification table as entity nodes includes: The association between entity nodes represented by fields in the same row of document is determined according to the preset inter-column relationship mapping rules, and is used as the corresponding entity node relationship; Construct a knowledge subgraph based on each entity node and its corresponding relationship with other entity nodes.

[0008] In one possible implementation, the step of performing similarity comparison on entity nodes in each knowledge subgraph and merging the knowledge subgraphs based on the comparison results to obtain the corresponding initial knowledge graph includes: The first entity node set is determined based on the first entity node of each knowledge subgraph. The first entity node set includes the functional unit node set, the fault mode node set, and the fault phenomenon node set. For a set of entities of the same type in the first set of entity nodes, calculate the similarity between any two entity nodes in the set to obtain the first similarity comparison result; Entity nodes whose first similarity comparison results exceed a preset similarity threshold are merged into a representative entity node; The second entity node set is determined based on the second entity node of each knowledge subgraph. The second entity node set includes the fault cause node set and the fault impact node set. For each entity node in the second entity node set, the similarity between the entity node and the representative entity node of each failure mode is calculated to obtain the second similarity comparison result. Entity nodes whose second similarity comparison results exceed the preset similarity threshold are merged, and the corresponding entity node relationships are synchronized to obtain the corresponding initial knowledge graph.

[0009] In one possible implementation, after performing similarity comparison on entity nodes in each knowledge subgraph and merging the knowledge subgraphs based on the comparison results to obtain the corresponding initial knowledge graph, the method further includes: Determine the target entity nodes in the initial knowledge graph, wherein the target entity nodes include representative entity nodes of the fault modes that have no entity node relationship with the functional unit nodes in the initial knowledge graph; For each target entity node, perform word segmentation to obtain the target semantic phrase of the corresponding word segmentation window; For each target semantic phrase, the similarity between the phrase and each functional unit node in the initial knowledge graph is calculated to obtain the third similarity comparison result. The nodes whose third similarity comparison results exceed the preset similarity threshold are supplemented with their corresponding attribution relationships to adjust the initial knowledge graph.

[0010] In one possible implementation, determining the target fault mode node corresponding to the fault mode segment in the initial knowledge graph, and supplementing the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph, includes: For each fault pattern segment, word segmentation is performed to obtain the fault pattern semantic phrase of the corresponding word segmentation window; For each fault mode semantic phrase, calculate the similarity between the phrase and each fault mode node in the initial knowledge graph to obtain the fourth similarity comparison result; The fault mode node whose fourth similarity comparison result exceeds the preset similarity threshold is taken as the target fault mode node; Each maintenance operation segment is preprocessed, and the preprocessed maintenance operation segments are used as maintenance nodes. The maintenance node is associated with the target fault mode node in the initial knowledge graph to obtain the corresponding target operation and maintenance management knowledge graph.

[0011] In one possible implementation, the step of determining the target fault mode node corresponding to the fault mode segment in the initial knowledge graph, and supplementing the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph, further includes: For each fault mode semantic phrase, determine the degree of overlap between the phrase and the other fault mode semantic phrases, and retain fault mode semantic phrases whose degree of overlap exceeds a set threshold.

[0012] A second aspect of this disclosure provides a knowledge graph construction apparatus for operation and maintenance management, the apparatus comprising: The data acquisition module is used to acquire fault management specification documents and maintenance management specification documents; The cleaning module is used to extract corresponding document fields from the fault management specification file according to the preset content ontology, and to construct a fault management specification table based on the content ontology and the document fields; The knowledge subgraph construction module is used to perform structural transformation on the fault management specification table, and construct a corresponding knowledge subgraph by taking the document field of each row of the fault management specification table as the corresponding entity node. The entity node includes at least the fault mode node. The initial knowledge graph construction module is used to compare the similarity of entity nodes in each knowledge subgraph and merge the knowledge subgraphs according to the comparison results to obtain the corresponding initial knowledge graph. The maintenance data extraction module is used to extract fault mode segments and matching maintenance operation segments from the maintenance management specification document; The target result generation module is used to determine the target fault mode node in the initial knowledge graph that corresponds to the fault mode segment, and to complete the target fault mode node with the maintenance operation segment as a maintenance node, so as to obtain the target operation and maintenance management knowledge graph.

[0013] A third aspect of this disclosure provides an electronic device comprising: At least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the knowledge graph construction method for operation and maintenance management as described in this disclosure.

[0014] A fourth aspect of this disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the knowledge graph construction method for operation and maintenance management described in this disclosure.

[0015] This disclosure discloses a method for constructing a knowledge graph for operations and maintenance (O&M) health management. Based on fault management and maintenance management specifications, a corresponding target knowledge graph is built. Specifically, by pre-setting a content ontology, corresponding document fields are extracted from the fault management specifications to determine entity nodes in the graph. Specifications from different vendors can all have their corresponding fields determined based on the content ontology, enabling rapid alignment of multi-source data. Furthermore, based on the knowledge graph constructed from the document fields extracted from the fault management specifications, corresponding maintenance nodes are added. This ensures that the final target O&M management knowledge graph can not only locate fault modes but also directly determine corresponding maintenance methods, providing reliable data support for O&M management and improving the efficiency and quality of handling sudden abnormal events in various O&M management scenarios.

[0016] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0017] The above and other objects, features, and advantages of this disclosure will become readily apparent from the following detailed description of exemplary embodiments, taken in conjunction with the accompanying drawings. Several embodiments of this disclosure are illustrated in the drawings by way of example and not limitation, in which: In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0018] Figure 1 The illustration shows a schematic diagram of the implementation process of a knowledge graph construction method for operation and maintenance health management according to an embodiment of the present disclosure; Figure 2 A schematic diagram of a knowledge subgraph structure according to an embodiment of this disclosure is shown; Figure 3 This illustration shows a schematic diagram of an initial knowledge graph structure according to an embodiment of the present disclosure; Figure 4 This illustration shows a schematic diagram of an initial knowledge graph after supplementing attribution relationships according to an embodiment of the present disclosure; Figure 5 This disclosure illustrates a target operation and maintenance management knowledge graph according to an embodiment of the present disclosure; Figure 6 This illustration shows a schematic diagram of a knowledge graph construction device for operation and maintenance health management according to an embodiment of the present disclosure; Figure 7 A schematic diagram of the composition structure of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0019] To make the objectives, features, and advantages of this disclosure more apparent and understandable, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0020] Knowledge graphs, as a novel management system for documents, can define knowledge dimensions that transcend the documents themselves, thereby forming an integrated knowledge management capability for dispersed data. However, existing health management knowledge graphs are still quite separate from the objective needs of actual operation and maintenance scenarios, and a large amount of documents still need to be manually entered, resulting in significant shortcomings in terms of application adaptability and cost.

[0021] The main drawbacks of the relevant technologies are as follows: On the one hand, the traditional knowledge graph construction process requires a lot of manual intervention, which is difficult to meet the increasingly large scale of equipment and failures in scenarios such as airports, and has significant application bottlenecks; on the other hand, the knowledge graph construction method based on language large models not only relies on a lot of computing resources, but also often has the objective risk of the granularity of the mined content not being consistent with the actual knowledge application needs, and still has significant application defects.

[0022] On the other hand, the construction goals of health management knowledge graphs are not clear enough, and they are not closely aligned with actual operation and maintenance processes. They unilaterally emphasize the general knowledge mining capabilities of large language models, while ignoring the structured professional knowledge characteristics of materials such as Failure Mode and Effects Analysis (FMEA) documents. This results in a complex and chaotic knowledge system, which seriously affects subsequent fault location and handling based on knowledge graphs.

[0023] Based on this, this disclosure provides a method for constructing a knowledge graph for operation and maintenance health management to solve the above problems, such as... Figure 1 As shown, the method includes: S101. Obtain the fault management specification document and the maintenance management specification document; In this step, the fault management specification documents include the specification FMEA document, and the maintenance management specification documents are the corresponding maintenance manuals used for operation and maintenance management. These serve as the data foundation for building the corresponding knowledge graph.

[0024] S102. Extract the corresponding document fields from the fault management specification document according to the preset content ontology, and construct the fault management specification table according to the content ontology and document fields; In this step, the content ontology is defined based on the actual business characteristics of operations and maintenance health management, including functional units, failure modes, failure phenomena, failure causes, local impact, superior impact, and ultimate impact. Correspondingly, based on the above ontology content, the corresponding document fields are extracted from the failure management specification documents to obtain the corresponding failure management specification table. It should be noted that the styles of specification FMEA documents from different vendors vary greatly, especially in writing style or column name definitions. Although the documents from different vendors are complete in content, their inconsistent formats make it difficult to directly apply them to the construction of a knowledge graph that reflects the actual situation. Therefore, this step pre-defines the content ontology and determines the document fields in each FMEA document based on the content ontology for subsequent knowledge graph construction.

[0025] S103. Transform the fault management specification table into a structure, and use the document fields of each row in the fault management specification table as corresponding entity nodes to construct the corresponding knowledge subgraph. The entity nodes shall include at least the fault mode nodes. In this step, the document fields corresponding to the content ontology are identified, resulting in a fault management specification table. The column names in the fault management specification table correspond to the content ontology, and the content of each row corresponds to the document fields determined based on the content ontology. For the fault specification table, a knowledge subgraph is constructed for each row, using the document fields of the same row as entity nodes. This step, based on actual business characteristics and centered on fault patterns, ensures that each knowledge subgraph includes at least the nodes corresponding to the fault pattern, resulting in multiple knowledge subgraphs.

[0026] S104. Perform similarity comparison on the entity nodes in each knowledge subgraph, and merge the knowledge subgraphs according to the comparison results to obtain the corresponding initial knowledge graph. In this step, for all knowledge subgraphs, the similarity between any two entity nodes is calculated, thereby merging two entity nodes whose similarity meets the similarity threshold, so as to merge multiple knowledge subgraphs into an initial knowledge graph and reduce the occurrence of redundant nodes with the same or similar meanings in the initial knowledge graph.

[0027] S105. Extract the fault mode phrases and matching maintenance operation phrases from the maintenance management specification documents; In this step, the obtained maintenance management specification document, namely the maintenance manual, is used to extract fault mode segments and maintenance segments that match the fault mode segments. This serves as the data foundation to further supplement the initial knowledge graph obtained in step S104. Based on the initial knowledge graph, a graph that can determine the corresponding maintenance methods for faults is obtained, thereby further enhancing the graph's functionality.

[0028] S106. Determine the target fault mode node corresponding to the fault mode segment in the initial knowledge graph, and complete the target fault mode node with the maintenance operation segment as the maintenance node to obtain the target operation and maintenance management knowledge graph.

[0029] In this step, by analyzing the fault mode segment, the fault mode node corresponding to the segment in the initial knowledge graph is determined, which is the target fault mode node. Then, the maintenance node represented by the processed maintenance operation segment can be supplemented to the target fault mode node based on the association with the fault segment, thereby obtaining a complete target operation and maintenance management knowledge graph.

[0030] This disclosure provides a method for constructing a knowledge graph for operations and maintenance (O&M) health management. Based on fault management and maintenance management specifications, a corresponding target knowledge graph is constructed. Specifically, corresponding document fields are extracted from the fault management specifications using a pre-defined content ontology to determine entity nodes in the graph. Specifications from different vendors can all have their corresponding fields determined based on the content ontology, enabling rapid alignment of multi-source data. Furthermore, in addition to the knowledge graph constructed from the document fields extracted from the fault management specifications, corresponding maintenance nodes are added. This ensures that the final target O&M management knowledge graph can not only locate fault modes but also directly determine corresponding maintenance methods, providing reliable data support for O&M management, effectively reducing human intervention, and improving the efficiency and quality of handling sudden abnormal events in various O&M management scenarios.

[0031] In one possible implementation, corresponding document fields are extracted from the fault management specification document based on a preset content ontology, and a fault management specification table is constructed based on the content ontology and document fields, including: Determine the corresponding column name fields in the fault management specification file based on the preset content ontology; Construct the corresponding ontology template based on the content ontology and column name fields; Identify the document fields corresponding to the column name fields in the fault management specification file, and replace the column name fields in the ontology template with the document fields to obtain the corresponding fault management specification table.

[0032] In this embodiment, when processing the fault management specification document, the corresponding column name field in the specification document is first determined based on the preset content ontology, as shown in Table 1. Table 1 shows a portion of the content of a specification document.

[0033]

[0034] Table 1. Suggested Contents of the Standard Document It should be noted that this embodiment determines the most suitable column name for the content ontology by performing sentence vector encoding on the column names in both the content ontology and the fault management specification file. Accordingly, each content ontology—namely, functional unit, fault mode, fault phenomenon, fault cause, local impact, superior impact, and final impact—as well as the column names in the FMEA file, are encoded using the Sentence-BERT sentence vector model. Then, based on the sentence vector encoding, a suitable column name is selected for each content ontology. The corresponding formula is as follows:

[0035] In the formula, and Representing the content ontology respectively With column names The sentence vector, The cosine similarity between sentence vectors Representative and Content Ontology The column name with the best cosine similarity to the sentence vector.

[0036] As shown in Table 1, the column names corresponding to the above content in this specification document are, in order: drawing number, fault mode, detection method, cause, local fault impact, higher-level fault impact, and final fault impact.

[0037] Furthermore, based on the content ontology and the selected column names, a corresponding ontology template is constructed, for example: { "Functional Unit": "{Drawing Number}", "Fault Mode": "{Fault Mode}", "Fault Phenomenon": "{Detection Method}", "Cause of Fault": "{Cause}", "Local impact": "{Local failure impact}", "Impact from higher authorities": "{Impact of a failure at the next higher level}", "Final Impact": "{Final Failure Impact}" } In other words, the ontology template uses "{}" to mark column name fields, thereby constructing a formal template for cleaning file content using strings as carriers, achieving rapid alignment of multi-source FMEA files. It should be noted that this embodiment further identifies the document fields corresponding to the aforementioned column names from the FMEA file. For example, as shown in Table 1, the document field with the column name "Figure Number" represents power supply or motor, etc. The identified document fields replace the original column names in the ontology template; for example, replacing the figure number in the ontology template with "Power Supply." This determines all corresponding document fields obtained from the FMEA file, resulting in the corresponding fault management specification table, which serves as the data foundation for subsequent knowledge graph construction. This embodiment highlights the core position of faults in the proactive health management knowledge system, constructing a knowledge structure organization paradigm that meets the objective needs of proactive health management business, making the resulting graph more aligned with operational requirements.

[0038] In one possible implementation, before determining the document field corresponding to the column name field in the fault management specification document, the method further includes: Obtain the corrected ontology template to determine the corresponding document fields based on the column name fields in the corrected ontology template.

[0039] In this embodiment, a verification step is added before determining the document fields corresponding to the FMEA file. That is, the constructed ontology template is returned to the user for verification, and the user further optimizes the template to obtain a corrected ontology template. Based on this, the document fields determined are more accurate.

[0040] The corrected ontology template is shown below: { "Functional Unit": "{Drawing Number}", "Fault Mode": "{Drawing Number}{Fault Mode}", "Fault Phenomenon": "{Drawing Number}{Detection Method}", "Cause of Fault": "{Cause}", "Local impact": "{Local failure impact}", "Impact from higher authorities": "{Impact of a failure at the next higher level}", "Final Impact": "{Final Failure Impact}" } Based on this, the document fields corresponding to the FMEA file are determined, and the resulting fault management specification table is shown in Table 2.

[0041]

[0042] Table 2 Fault Management Standard Table In other words, this embodiment returns the matched column name results to the user for verification in the format of "{matched column name field}". The user can then further optimize and modify the template. After the user completes the modification and submits the template, the corresponding column name information, i.e., the document field, is extracted according to the modified template. Using each content body as the column name, the information in each row is replaced with the "{}" part according to the content corresponding to the column name, thereby cleaning out the document information of the corresponding content body.

[0043] In one possible implementation, the document fields of each row in the fault management specification table are used as entity nodes to construct the corresponding knowledge subgraph, including: The association between entity nodes represented by fields in the same row of document is determined according to the preset inter-column relationship mapping rules, and is used as the corresponding entity node relationship; Construct a knowledge subgraph based on each entity node and its corresponding relationship with other entity nodes.

[0044] In this embodiment, several knowledge subgraphs are constructed based on the obtained fault management specification table. It should be noted that this embodiment performs structured extraction conforming to the specification knowledge graph architecture definition on the content of the fault management specification table. For each row of the fault management specification table, the content is extracted to construct a corresponding knowledge subgraph. For example, for the row in Table 2 where the functional unit is "power supply," the content is extracted, and each corresponding content node is used to construct a knowledge subgraph for that row. Similarly, the content for the row where the functional unit is "motor" is also extracted, and a corresponding knowledge subgraph is constructed, such as... Figure 2 As shown.

[0045] It should be noted that the relationships between entity nodes in each knowledge subgraph are determined based on preset inter-column relationship mapping rules. For relationships that can be directly resolved through structural analysis, they are directly constructed based on information within the same data row. These relationships include fault modes belonging to functional units, fault causes leading to fault modes, fault modes causing local impacts, local impacts causing upper-level impacts, and upper-level impacts causing final impacts. There are clear inter-column relationships in the FMEA file, which can be determined by analyzing the information structure of the same data row, as shown in Table 3.

[0046]

[0047] Table 3 Entity Relationship Table 1 Therefore, in the knowledge subgraph, the entity node relationship between the fault mode (source entity) and the functional unit (target entity) is that the fault mode belongs to the functional unit, that is, the fault mode node points to the functional unit node. The other relationships are shown in Table 3, and the repeated parts will not be repeated.

[0048] In one implementation, similarity comparisons are performed on entity nodes in each knowledge subgraph, and the knowledge subgraphs are merged based on the comparison results to obtain the corresponding initial knowledge graph, including: The first entity node set is determined based on the first entity node of each knowledge subgraph. The first entity node set includes the functional unit node set, the fault mode node set, and the fault phenomenon node set. For a set of entities of the same type in the first set of entity nodes, calculate the similarity between any two entity nodes in the set to obtain the first similarity comparison result; Entity nodes whose first similarity comparison results exceed a preset similarity threshold are merged into a representative entity node; The set of second entity nodes is determined based on the second entity node of each knowledge subgraph. The set of second entity nodes includes the set of fault cause nodes and the set of fault impact nodes. For each entity node in the second entity node set, calculate the similarity between the entity node and the representative entity node of each failure mode to obtain the second similarity comparison result; Entity nodes whose second similarity comparison results exceed the preset similarity threshold are merged, and the corresponding entity node relationships are synchronized to obtain the corresponding initial knowledge graph.

[0049] In this embodiment, for all entity nodes in the knowledge subgraph, similar nodes are merged by determining the similarity between nodes. Accordingly, for all knowledge subgraphs, a corresponding first entity node set is determined. It should be noted that, in this embodiment, the first entity nodes of the knowledge subgraph include functional unit nodes, fault mode nodes, and fault phenomenon nodes. The corresponding first entity node sets include functional unit node sets, fault mode node sets, and fault phenomenon node sets, that is, three node sets are constructed, and each node set involves only nodes of the same type.

[0050] Furthermore, for each set of nodes of the same type, the similarity between any two entity nodes within the set is calculated. For example, for a set of functional unit nodes, the cosine similarity between any two functional unit nodes within the set is calculated to obtain the corresponding first similarity comparison result. If this result exceeds a preset similarity threshold, for example, if the similarity comparison result between two functional unit nodes exceeds 70%, then these two functional unit nodes are merged into a single representative node for a functional unit. Similarly, two fault mode functional units are merged into a single representative node for a fault mode, and so on, thereby merging the various knowledge subgraphs.

[0051] It should also be noted that the second entity nodes of the knowledge subgraph include fault cause nodes and fault impact nodes. Among them, fault impact nodes include local impact nodes, superior impact nodes, and final impact nodes. Accordingly, based on the second entity nodes of all knowledge subgraphs, a corresponding set of second entity nodes is determined, and each entity node in the set is compared with the representative node of each fault mode to obtain the second similarity comparison result. The fault cause nodes, local impact nodes, superior impact nodes, and final impact nodes are then loaded into the corresponding fault mode nodes one by one. The specific embedding relationship is shown in Table 4.

[0052]

[0053] Table 4. Embedding Relationship Comparison Table It should be noted that, based on the second similarity comparison results, the second entity nodes with a sentence vector similarity exceeding 75% are merged with the corresponding representative nodes of the fault mode. This process is repeated to further remove redundant nodes from the graph, resulting in the corresponding initial knowledge graph, such as... Figure 3As shown, the original entity node relationships corresponding to each node are automatically associated with the corresponding node. That is, this embodiment, based on multiple types of entities, realizes the autonomous and standardized mining of fault propagation logic through multi-step mining and organization methods. The constructed graph can effectively improve the retrieval and analysis efficiency in subsequent applications.

[0054] In one possible implementation, after performing similarity comparisons on entity nodes in each knowledge subgraph and merging the knowledge subgraphs based on the comparison results to obtain the corresponding initial knowledge graph, the method further includes: Identify target entity nodes in the initial knowledge graph. Target entity nodes include representative entity nodes of fault modes that do not have any entity node relationship with functional unit nodes in the initial knowledge graph. For each target entity node, perform word segmentation to obtain the target semantic phrase of the corresponding word segmentation window; For each target semantic phrase, the similarity between the phrase and each functional unit node in the initial knowledge graph is calculated to obtain the third similarity comparison result. Nodes whose third similarity comparison results exceed the preset similarity threshold are supplemented with their corresponding attribution relationships to adjust the initial knowledge graph.

[0055] In this embodiment, for the merged initial knowledge graph, representative entity nodes of fault modes that do not have any entity node relationship with functional unit nodes are identified, i.e., target entity nodes. These include local influence nodes of the fault mode nodes, parent influence nodes of the fault mode nodes, and final influence nodes of the fault mode nodes. For the above nodes, there may be cases where they do not match the functional units in the same data row. For example, in the data row with the figure number "power supply" in Table 2, the corresponding local influence content is "low motor torque," meaning that the actual functional unit of this local influence is the motor, which does not match the power supply functional unit in this data row. The fault of low motor torque does not belong to the power supply fault. The corresponding association relationship cannot be determined directly based on the structural analysis of the same data row, so further semantic analysis is required. That is, some associations are already indicated by different entities belonging to the same entry in the standard table, so they can be directly analyzed through structural analysis; while some associations are difficult to indicate through the structural information of the standard table and need to be matched with corresponding objects through semantic analysis. Among them, the association mining based on semantic analysis needs to be performed after all structural analysis is completed. The specific association relationships and analysis methods are shown in Table 5.

[0056]

[0057] Table 5 Entity Relationship Table 2 As shown in Table 5, semantic analysis is performed on each target entity node. This is done through sentence vector mining and word segmentation using the Jieba analysis tool, forming target semantic phrases for corresponding word segmentation windows, such as semantic phrases for word segmentation windows 1 to 5. The similarity of these semantic phrases with each functional unit node is compared to obtain the third similarity comparison result. Based on the third similarity comparison result, the functional units with the highest sentence vector similarity and those with a cosine similarity exceeding 75% are selected as candidate functional units. Furthermore, the functional unit nodes with the highest phrase segmentation window and the deepest functional unit structure are selected as target functional unit nodes. This supplements the attribution relationship between target entity nodes and target functional unit nodes, for example... Figure 4 As shown, the relationship between power supply aging and power supply hierarchy is completed. Among all matched functional units, the smallest unit (the lowest level unit) is the functional unit with the deepest structure. For example, in the functional units "Power System" and "Power System Battery Pack," "Power System Battery Pack" is a component of "Power System." Accordingly, the structure of the "Power System Battery Pack" functional unit is smaller and deeper than that of the "Power System" functional unit.

[0058] In one implementation, the target fault mode node corresponding to the fault mode segment in the initial knowledge graph is determined, and the maintenance operation segment is added as a maintenance node to the target fault mode node to obtain the target operation and maintenance management knowledge graph, including: For each fault pattern segment, word segmentation is performed to obtain the fault pattern semantic phrase of the corresponding word segmentation window; For each fault mode semantic phrase, calculate the similarity between the phrase and each fault mode node in the initial knowledge graph to obtain the fourth similarity comparison result. The fault mode node whose fourth similarity comparison result exceeds the preset similarity threshold is taken as the target fault mode node; Each maintenance operation segment is preprocessed, and the preprocessed maintenance operation segments are used as maintenance nodes. By establishing associations between maintenance nodes and target fault mode nodes in the initial knowledge graph, the corresponding target operation and maintenance management knowledge graph is obtained.

[0059] In this embodiment, based on the initial knowledge graph constructed from the fault management specification document, maintenance-related proactive health management knowledge is further supplemented. It should be noted that the fault mode nodes in the initial knowledge graph serve as carriers for supplementing the graph with relevant maintenance knowledge.

[0060] It should be noted that the maintenance management specification document in this embodiment must meet the preset writing specifications, including: all maintenance operations must first declare the applicable fault mode, "." is not allowed during the enumeration of multiple fault modes, and specific maintenance steps are connected with ":", and enumerated with numbers. "." is prohibited before the enumeration is completed.

[0061] When extracting fault mode phrases and matching maintenance operation phrases from maintenance management specification documents, search for all colons (:) in the document, and trace back to the beginning of each colon (.") as the fault mode phrase, and trace back to the end of the text after the colon (.") as the maintenance operation phrase. For example, a portion of a maintenance management specification document is shown below: "Power supply batteries are highly susceptible to aging. If battery aging occurs, then:" 1. Disconnect the power supply.

[0062] 2. Remove the battery.

[0063] 3. Install the new battery.

[0064] 4. Keep the power supply connected.

[0065] Processing procedure: For the above repair content, search for all colons (:), and trace back to the beginning of each colon (.") as the fault mode segment, and trace back to the end of the text (.") as the repair operation segment. The corresponding fault mode segment is "If battery aging occurs, then", and the repair operation segment is "1. Disconnect the power. 2. Remove the battery. 3. Install a new battery. 4. Reconnect the power."

[0066] Furthermore, the aforementioned fault pattern segments are segmented using the Jieba word segmentation tool to form semantic phrases with 1 to 5 segmentation windows, for example: ['If', 'occur', 'battery', 'aging', 'phenomenon', ',', 'then', 'If', 'battery', 'battery aging', 'aging phenomenon', 'phenomenon,', ',then', 'If battery', 'battery aging', 'battery aging phenomenon', 'aging phenomenon,', 'phenomenon,then', 'If battery aging', 'battery aging phenomenon', 'battery aging phenomenon,', 'aging phenomenon,then', 'If battery aging phenomenon', 'battery aging phenomenon,', 'battery aging phenomenon,then'。]

[0067] The sentence vectors of each phrase are compared with the nodes of each fault mode (fault modes have representative nodes) to obtain the fourth similarity comparison results, for example: ('If', 'Battery aging') = 0.1% ('Battery aging', 'Battery aging') = 100% Then, the fault pattern node with the highest cosine similarity to the phrase vector of each phrase and a cosine similarity of more than 75% is selected as a candidate. If there is no fault pattern node with a cosine similarity of more than 75%, all matching results of the phrase are discarded in order to determine the target fault pattern node.

[0068] Furthermore, the above maintenance segments are preprocessed. Accordingly, they are segmented according to the regular expression "[0-9a-zA-ZivxIVX\+\-`]+[ \t\.、]+?", and the spaces and newlines at the beginning and end of each maintenance segment are removed. This way, the non-empty maintenance operations are used as the set of maintenance operations for each maintenance segment, which serves as the basis for maintenance nodes. In the [0-9a-zA-ZivxIVX\+\-`] section, [...] represents a "character group," matching any single character within the brackets; 0-9: Arabic numerals (e.g., steps 1, 2, 3…); ivxIVX: Roman numerals (e.g., steps i, ii, iii, IV, V…, adapting to Roman numeral numbers in some manuals); \+\-: escaped "+" and "-" (matching prefixes such as "3+" and "2-" that may appear in the steps); `: backticks (adapting to steps marked with backticks in some manuals, such as "1. Check the power"); In the [ \t\.、]+? section, []: character class, matching the "separator between the sequence number and the operation" within the brackets; \t: tab character (a space generated by the Tab key; some manuals use Tab indentation to separate the sequence number and the operation); \.: escaped period ("." is a special symbol in regular expressions that "matches any character," requiring escape with "\", matching "1."). The regular expression removes the period (.) after the sequence number in "Disconnect the power"; 、: Chinese comma (to adapt to formats like "i、Check the circuit" in Chinese manuals that use commas for separation); +?: non-greedy quantifier, where "+" means "match one or more times" and "?" means "non-greedy matching" (matching as little as possible to avoid misinterpreting symbols in subsequent operation content as separators). In other words, based on the above regular expression, it precisely removes "sequence number / prefix + separator" from the step-by-step sections of the repair manual, retaining only the pure repair operation content.

[0069] For example: ["Disconnect power", "Remove battery", "Install new battery", "Reconnect power"].

[0070] In addition, entities with a similarity exceeding 75% are merged, and similar expressions are combined into a single word, ultimately yielding all maintenance nodes used to add to the initial knowledge graph. Since maintenance operation segments match fault mode segments, the relationship between maintenance nodes and corresponding fault mode nodes can be determined. Based on this, the initial knowledge graph is completed, resulting in the final target knowledge graph for operation and maintenance management. The resulting graph not only reflects actual fault modes but also simultaneously provides corresponding maintenance operations, ensuring the efficiency of on-site operation and maintenance management.

[0071] In one possible implementation, the target fault mode node corresponding to the fault mode segment in the initial knowledge graph is determined, and the maintenance operation segment is added as a maintenance node to the target fault mode node to obtain the target operation and maintenance management knowledge graph, which also includes: For each fault mode semantic phrase, determine the degree of overlap between the phrase and the other fault mode semantic phrases, and retain fault mode semantic phrases with an overlap degree exceeding a set threshold.

[0072] In this embodiment, each fault mode semantic phrase is analyzed one by one according to the segmentation window from largest to smallest. If the window of the current phrase can be covered by the window of another already processed phrase, the phrase is directly omitted. If the number of overlapping segments between the window of the current phrase and the window of another already processed phrase exceeds 25% of its own number of segments, the phrase is also omitted. That is, the degree of overlap between the current phrase and other fault mode semantic phrases determines whether to discard or retain the current phrase. Only fault mode semantic phrases with an overlap of less than 25% are retained, thereby determining all fault mode nodes of the current maintenance segment.

[0073] The main objective of this embodiment is to group terms such as "battery old" and "battery aging" into the longest term "battery aging," and then match them to the representative entity nodes of the fault mode corresponding to "battery aging" on the initial knowledge graph. The resulting target knowledge graph is as follows: Figure 5 As shown, the target knowledge graph includes representative entity nodes for fault modes, functional units, and fault phenomena, as well as maintenance measure nodes. Correspondingly, the relationship between maintenance measure nodes and fault mode nodes includes: a triggering relationship between fault mode nodes and each maintenance measure node, representing the maintenance measure corresponding to the fault mode trigger; and a continuation relationship between maintenance measure nodes based on the order of the corresponding maintenance segments. For example, the maintenance measures triggered by the battery aging fault mode include disconnecting the power, continuing to disassemble the battery, further installing a new battery, and then further reconnecting the power. The description of each node in the target knowledge graph is shown in Table 6.

[0074]

[0075] Table 6. Schematic diagram of entity node types The corresponding entity node relationships are shown in Table 7.

[0076]

[0077] Table 7. Schematic diagram of entity relationships In summary, the method disclosed herein constructs a corresponding knowledge graph based on fault management and maintenance management specification documents. The resulting graph uses fault modes as the core ontology, including fault phenomena, functional units, and maintenance measures, among other types of nodes (secondary ontology). All other types of nodes are aligned with fault modes during the construction process to ensure the accuracy and effectiveness of the graph. On the one hand, a lightweight analysis method for health management knowledge is achieved through a professional knowledge mining mechanism based on sentence vectors, combined with the semantic analysis capabilities of sentence vectors. On the other hand, a fault-based health management knowledge framework is constructed, fully utilizing the inherent knowledge structuring advantages of formalized materials such as FMEAs and maintenance manuals. This enables standardized management of multi-source health management knowledge for actual operation and maintenance scenarios. In the face of dispersed and complex operation and maintenance scenarios, such as airport operation and maintenance scenarios, this method can effectively improve the efficiency and quality of handling abnormal events. Specifically, by effectively integrating all necessary entities and their relationships through the structured parsing of the standardized FMEA documents and the ontology recognition of the maintenance manuals, multi-source documents are structured into a unified and clear standardized knowledge graph, effectively supporting the efficient retrieval and analysis needs of knowledge in the subsequent proactive health management phase.

[0078] To implement the above method, an example of this application also provides a knowledge graph construction device 600 for operation and maintenance health management, such as... Figure 6 As shown, the device includes: Data acquisition module 601 is used to acquire fault management specification documents and maintenance management specification documents; The cleaning module 602 is used to extract corresponding document fields from the fault management specification file according to the preset content ontology, and to construct a fault management specification table according to the content ontology and the document fields; The knowledge subgraph construction module 603 is used to perform structural transformation on the fault management specification table, and construct a corresponding knowledge subgraph by taking the document field of each row of the fault management specification table as the corresponding entity node. The entity node includes at least a fault mode node. The initial knowledge graph construction module 604 is used to perform similarity comparison on entity nodes in each knowledge subgraph and merge the knowledge subgraphs according to the comparison results to obtain the corresponding initial knowledge graph. The maintenance data extraction module 605 is used to extract fault mode segments and matching maintenance operation segments from the maintenance management specification document; The target result generation module 606 is used to determine the target fault mode node in the initial knowledge graph that corresponds to the fault mode segment, and to complete the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph.

[0079] By way of example, this application also provides an electronic device, including: processor; Memory used to store processor-executable instructions; The processor is used to read executable instructions from memory and execute the instructions to implement the knowledge graph construction method for operation and maintenance health management described above.

[0080] For example, this application also provides a computer-readable storage medium storing a computer program for executing the above-described knowledge graph construction method for operation and maintenance health management.

[0081] Figure 7 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0082] like Figure 7 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 402 or a computer program loaded from storage unit 408 into random access memory (RAM) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0083] Multiple components in device 1000 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of monitors, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0084] The computing unit 401 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as a knowledge graph construction method for operational health management. For example, in some embodiments, a knowledge graph construction method for operational health management can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of the knowledge graph construction method for operational health management described above can be performed. Alternatively, in other embodiments, computing unit 401 may be configured by any other suitable means (e.g., by means of firmware) to perform a knowledge graph construction method for operational health management.

[0085] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0086] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0087] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0088] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0089] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0090] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0091] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0092] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0093] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A method for constructing a knowledge graph for operation and maintenance health management, characterized in that, The method includes: Obtain the fault management specification document and the maintenance management specification document; Extract the corresponding document fields from the fault management specification file according to the preset content ontology, and construct the fault management specification table according to the content ontology and the document fields; The fault management specification table is structurally transformed, and the document fields of each row in the fault management specification table are used as corresponding entity nodes to construct a corresponding knowledge subgraph. The entity nodes include at least fault mode nodes. The similarity of entity nodes in each knowledge subgraph is compared, and the knowledge subgraphs are merged according to the comparison results to obtain the corresponding initial knowledge graph. Extract the fault mode phrases and matching maintenance operation phrases from the maintenance management specification document; Identify the target fault mode node in the initial knowledge graph that corresponds to the fault mode segment, and complete the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph.

2. The knowledge graph construction method for operation and maintenance health management according to claim 1, characterized in that, The step of extracting corresponding document fields from the fault management specification file according to a preset content ontology, and constructing a fault management specification table based on the content ontology and the document fields, includes: The corresponding column name fields in the fault management specification file are determined based on the preset content ontology; Construct a corresponding ontology template based on the content ontology and the column name field; Identify the document field corresponding to the column name field in the fault management specification file, and replace the column name field in the ontology template with the document field to obtain the corresponding fault management specification table.

3. The knowledge graph construction method for operation and maintenance health management according to claim 2, characterized in that, Before determining the document field corresponding to the column name field in the fault management specification file, the method further includes: Obtain the corrected ontology template, and determine the corresponding document fields based on the column name fields in the corrected ontology template.

4. The knowledge graph construction method for operation and maintenance health management according to claim 1, characterized in that, The step of constructing a corresponding knowledge subgraph by using the document fields of each row in the fault management specification table as entity nodes includes: The association between entity nodes represented by fields in the same row of document is determined according to the preset inter-column relationship mapping rules, and is used as the corresponding entity node relationship; Construct a knowledge subgraph based on each entity node and its corresponding relationship with other entity nodes.

5. The knowledge graph construction method for operation and maintenance health management according to claim 1, characterized in that, The process of performing similarity comparisons on entity nodes in each knowledge subgraph and merging the knowledge subgraphs based on the comparison results to obtain the corresponding initial knowledge graph includes: The first entity node set is determined based on the first entity node of each knowledge subgraph. The first entity node set includes the functional unit node set, the fault mode node set, and the fault phenomenon node set. For a set of entities of the same type in the first set of entity nodes, calculate the similarity between any two entity nodes in the set to obtain the first similarity comparison result; Entity nodes whose first similarity comparison results exceed a preset similarity threshold are merged into a representative entity node; The second entity node set is determined based on the second entity node of each knowledge subgraph. The second entity node set includes the fault cause node set and the fault impact node set. For each entity node in the second entity node set, the similarity between the entity node and the representative entity node of each failure mode is calculated to obtain the second similarity comparison result. Entity nodes whose second similarity comparison results exceed the preset similarity threshold are merged, and the corresponding entity node relationships are synchronized to obtain the corresponding initial knowledge graph.

6. The knowledge graph construction method for operation and maintenance health management according to claim 5, characterized in that, After performing similarity comparisons on entity nodes in each knowledge subgraph and merging the knowledge subgraphs based on the comparison results to obtain the corresponding initial knowledge graph, the method further includes: Determine the target entity nodes in the initial knowledge graph, wherein the target entity nodes include representative entity nodes of the fault modes that have no entity node relationship with the functional unit nodes in the initial knowledge graph; For each target entity node, perform word segmentation to obtain the target semantic phrase of the corresponding word segmentation window; For each target semantic phrase, the similarity between the phrase and each functional unit node in the initial knowledge graph is calculated to obtain the third similarity comparison result. The nodes whose third similarity comparison results exceed the preset similarity threshold are supplemented with their corresponding attribution relationships to adjust the initial knowledge graph.

7. The knowledge graph construction method for operation and maintenance health management according to any one of claims 1-6, characterized in that, The process of determining the target fault mode node corresponding to the fault mode segment in the initial knowledge graph, and supplementing the target fault mode node with the maintenance operation segment as a maintenance node, to obtain the target operation and maintenance management knowledge graph includes: For each fault pattern segment, word segmentation is performed to obtain the fault pattern semantic phrase of the corresponding word segmentation window; For each fault mode semantic phrase, calculate the similarity between the phrase and each fault mode node in the initial knowledge graph to obtain the fourth similarity comparison result; The fault mode node whose fourth similarity comparison result exceeds the preset similarity threshold is taken as the target fault mode node; Each maintenance operation segment is preprocessed, and the preprocessed maintenance operation segments are used as maintenance nodes. The maintenance node is associated with the target fault mode node in the initial knowledge graph to obtain the corresponding target operation and maintenance management knowledge graph.

8. The knowledge graph construction method for operation and maintenance health management according to claim 7, characterized in that, The step of determining the target fault mode node corresponding to the fault mode segment in the initial knowledge graph, and completing the target fault mode node with the maintenance operation segment as a maintenance node to obtain the target operation and maintenance management knowledge graph, further includes: For each fault mode semantic phrase, determine the degree of overlap between the phrase and the other fault mode semantic phrases, and retain fault mode semantic phrases whose degree of overlap exceeds a set threshold.

9. A knowledge graph construction device for operation and maintenance health management, characterized in that, The device includes: The data acquisition module is used to acquire fault management specification documents and maintenance management specification documents; The cleaning module is used to extract corresponding document fields from the fault management specification file according to the preset content ontology, and to construct a fault management specification table based on the content ontology and the document fields; The knowledge subgraph construction module is used to perform structural transformation on the fault management specification table, and construct a corresponding knowledge subgraph by taking the document field of each row of the fault management specification table as the corresponding entity node. The entity node includes at least the fault mode node. The initial knowledge graph construction module is used to compare the similarity of entity nodes in each knowledge subgraph and merge the knowledge subgraphs according to the comparison results to obtain the corresponding initial knowledge graph. The maintenance data extraction module is used to extract fault mode segments and matching maintenance operation segments from the maintenance management specification document; The target result generation module is used to determine the target fault mode node in the initial knowledge graph that corresponds to the fault mode segment, and to complete the target fault mode node with the maintenance operation segment as a maintenance node, so as to obtain the target operation and maintenance management knowledge graph.

10. An electronic device, characterized in that, include: At least one processor; The at least one processor is connected to a memory; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the knowledge graph construction method for operational health management as described in any one of claims 1-8.

Images