Knowledge graph-based unified fault search method for specialized vehicle device

Abstract

Disclosed in the present invention is a knowledge graph-based unified fault search method for a specialized vehicle device. The method comprises: acquiring device maintenance related data, and taking a device as a core to establish association relationships between the device maintenance related data; acquiring a fault report, and parsing the fault report to obtain parsing result data; constructing a device fault graph on the basis of the device maintenance related data and the parsing result data; integrating the device fault graph into a knowledge graph platform; receiving a search term entered by a current user in a search box, acquiring search results corresponding to the search term, and displaying the search results on a search result page, wherein the search results comprise a result list, data classification statistics, search result sorting, a knowledge cloud picture, and associated recommendations; and providing in the search result page an operation entry for knowledge exploration, and viewing related knowledge exploration of the current knowledge by the user by triggering the operation entry for knowledge exploration. The method solves the problem in the prior art that the search results are not accurate enough.

Description

Vehicle professional equipment fault unified retrieval method based on knowledge graph TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge graph, in particular to a vehicle professional equipment fault unified retrieval method based on knowledge graph. BACKGROUND

[0002] With the development of science and technology and economy, rail transit has become an essential means of transportation in people's daily life and national economic development, of course, the safe operation of rail transit has also been paid more and more attention. In the operation of rail transit, the performance of train key devices or components is directly related to the safety and reliability of vehicle operation.

[0003] When the existing vehicle professional equipment occurs an unusual fault, the repairman needs to retrieve similar faults from the fault case library, determine the specific cause of the fault, and then repair it. However, the current retrieval result from the fault case library is too wide, and it cannot accurately retrieve similar or identical cases, which is not conducive to the analysis of the fault by the repairman.

[0004] Therefore, there is an urgent need for a vehicle professional equipment fault unified retrieval method based on knowledge graph to provide online help for repairmen and improve their fault diagnosis and processing capabilities.

[0005] SUMMARY

[0006] The purpose of the embodiment of the present application is to provide a vehicle professional equipment fault unified retrieval method based on knowledge graph, which solves the problem of inaccurate retrieval results in the prior art.

[0007] To achieve the above purpose, the embodiment of the present application provides a vehicle professional equipment fault unified retrieval method based on knowledge graph, which specifically comprises:

[0008] Obtain equipment repair related data, take the equipment as the core, and establish the association relationship between the equipment repair related data, wherein the equipment repair related data includes asset data, location data, personnel data, work order data, fault code data, professional information data and equipment subsystem code data;

[0009] Obtain a fault report, analyze the fault report to obtain analysis result data, wherein the analysis result data includes the influence relationship between the equipment and the equipment, the association relationship between the equipment and the fault, and the cause-effect relationship between the fault and the fault;

[0010] Construct an equipment fault graph based on the equipment repair related data and the analysis result data;

[0011] Integrate the device fault atlas into a knowledge graph platform;

[0012] Receive a search word input by a current user in a search box, obtain a search result corresponding to the search word, and display the search result on a search result page, wherein the search result includes a result list, data classification statistics, search result sorting, a knowledge cloud diagram, and associated recommendations;

[0013] An operation entry for knowledge exploration is provided in the search result page, and a user views relevant knowledge exploration of current knowledge by triggering the operation entry for knowledge exploration.

[0014] On the basis of the above technical solution, the application can also be improved as follows:

[0015] Further, the associated relationship includes a parent-child relationship between devices, an influence relationship between devices, an association relationship between a fault and a device, a cause-effect relationship between faults, a cause-effect relationship between a fault phenomenon and a fault cause, a relationship between a fault and a fault phenomenon, a belonging relationship between a system and a specialty, a belonging relationship between a device and a system, a belonging relationship between a device and a line, a location of a device, a supply relationship between a device and a supplier, a belonging relationship between a fault and a fault report, an association relationship between a fault and an inspection operation, a reporting relationship between a fault report and a person, an auditing relationship between a fault report and a person, a belonging relationship between a person and a specialty, a belonging relationship between a person and a route, a belonging relationship between a station and a line, a relationship between a section and a line, a relationship between a main depot and a line, and a relationship between a vehicle depot and a line.

[0016] Further, the fault report is obtained, the fault report is analyzed to obtain analysis result data, including:

[0017] The fault report is subjected to corpus annotation, a knowledge extraction model is trained through corpus annotation, and a trained knowledge extraction model is obtained;

[0018] The fault report is subjected to knowledge extraction through the trained knowledge extraction model to obtain extraction data;

[0019] The extraction data and device maintenance related data are subjected to data fusion to obtain data fusion results;

[0020] The device fault atlas is adjusted based on the data fusion results.

[0021] Further, the device fault atlas is constructed based on the device maintenance related data and the analysis result data, including:

[0022] Vector values of all device names are calculated, and the vector values are saved to a vector library;

[0023] The fault report is analyzed to obtain a device name, and vector calculation is performed on the device name to obtain a device name vector value;

[0024] An equipment entity name is queried from a vector library through the device name vector value to the similarity within a threshold range;

[0025] It is judged whether the device name and the equipment entity name are correct, and if the device name and the equipment entity name are not correct, reselection is performed.

[0026] Further, the search term input by the current user in the search box is received, the search result corresponding to the search term is obtained, and the search result is displayed on the search result page, comprising:

[0027] The search term input by the current user is searched for an entity through the device fault graph to obtain a target entity;

[0028] The entity search includes prefix search, fuzzy search and accurate retrieval, and the accurate retrieval includes entity search through keyword and / or category and / or attribute constraint;

[0029] The center of the knowledge cloud map is switched based on the target entity, and the knowledge cloud map centered on the target entity is displayed.

[0030] Further, the search term input by the current user is searched for an entity through the device fault graph to obtain a target entity, comprising:

[0031] The number of associated entities corresponding to the target entity is set as a per-page display quantity threshold, and it is judged whether the number of associated entities is greater than the per-page display quantity threshold, and if so, the entities are paginated.

[0032] Further, the search term input by the current user in the search box is received, the search result corresponding to the search term is obtained, and the search result is displayed on the search result page, further comprising:

[0033] The search term is associated with the recommendation, and related equipment and related fault phenomenon are recommended based on the association relationship of the target entity.

[0034] Further, the search term input by the current user in the search box is received, the search result corresponding to the search term is obtained, and the search result is displayed on the search result page, comprising:

[0035] The knowledge cloud map is filtered again through secondary screening to generate a subgraph meeting the constraint condition, wherein the secondary screening condition includes subject attribute screening and relationship attribute screening.

[0036] Further, the vehicle professional equipment fault unified retrieval method based on the knowledge graph further comprises:

[0037] constructing a fault knowledge index, wherein the fault knowledge index comprises a classification drop-down box, search history records, equipment types, intent recognition, search association prompts, search hot lists, popular data, and high-frequency case recommendations;

[0038] searching for equipment information and fault information based on the fault knowledge index.

[0039] Further, the vehicle professional equipment fault unified retrieval method based on the knowledge graph further comprises:

[0040] in response to a knowledge card viewing request initiated by a user, displaying a knowledge card corresponding to a target entity.

[0041] The embodiments of the present application have the following advantages:

[0042] In the vehicle professional equipment fault unified retrieval method based on the knowledge graph, equipment repair related data is acquired, the equipment is taken as the core, and an association relationship between the equipment repair related data is established, wherein the equipment repair related data comprises asset data, location data, personnel data, work order data, fault code data, professional information data, and equipment subsystem code data; a fault report is acquired, the fault report is analyzed to obtain analysis result data, wherein the analysis result data comprises an influence relationship between equipment and equipment, an association relationship between equipment and faults, and a cause-effect relationship between faults and faults; a device fault graph is constructed based on the equipment repair related data and the analysis result data; the device fault graph is integrated into a knowledge graph platform; a search term input by a current user in a search box is received, a search result corresponding to the search term is acquired, and the search result is displayed on a search result page, wherein the search result comprises a result list, data classification statistics, search result sorting, a knowledge cloud map, and association recommendations; an operation entry for knowledge exploration is provided in the search result page, and a user views related knowledge exploration of current knowledge by triggering the operation entry for knowledge exploration, thereby solving the problem of inaccurate search results in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for those skilled in the art, other drawings can be derived from the provided drawings without creative labor.

[0044] FIG. 1 is a first flowchart of the vehicle professional equipment fault unified retrieval method based on the knowledge graph;

[0045] Fig. 2 is a data relationship diagram of the fault atlas of the equipment of the present application;

[0046] Fig. 3 is an architecture diagram of the vehicle professional equipment fault unified retrieval system based on the knowledge atlas of the present application;

[0047] Fig. 4 is an electronic device entity structure schematic diagram provided by the present application. DETAILED DESCRIPTION

[0048] The embodiments of the present application are described below by specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0049] Fig. 1 is a flow chart of the vehicle professional equipment fault unified retrieval method based on the knowledge atlas of the present application. As shown in Fig. 1, the vehicle professional equipment fault unified retrieval method based on the knowledge atlas provided by the embodiment of the present application includes the following steps:

[0050] S101, obtaining equipment maintenance related data, taking equipment as the core, establishing the association relationship between the equipment maintenance related data;

[0051] Specifically, the equipment maintenance related data includes asset data, location data, personnel data, work order data, fault code data, professional information data and equipment subsystem code data.

[0052] The asset data is used to store asset information such as equipment and components. The location data is used to store location information such as the warehouse location of the equipment. The personnel data is used to store personnel basic information. The work order data is used to store work order information, including the filling personnel, the planned time, the task content and the like. The fault code data is used to store fault code information. The professional information data is used to store professional basic information. The equipment subsystem code data is used to store subsystem basic information.

[0053] The work order data includes maintenance work order month down data and maintenance work order month up data. The maintenance work order month down data is used to store the month down maintenance work order information. The maintenance work order month up data is used to store the month up maintenance work order information.

[0054] The association relationship includes:

[0055] The parent-child relationship of the equipment and the equipment: the parent equipment of the equipment described in the LIMIS (lightweight production management system) can be used to construct the equipment tree to realize the local to global impact analysis combined with the equipment fault.

[0056] LIMIS is mainly used for production planning, tracking and data analysis, etc.

[0057] Device and device influence relationship: the mutual influence relationship when the device fails is extracted through the fault report, and the device fault correlation analysis can be carried out combined with the fault.

[0058] Fault and device correlation relationship: the device that produces fault is extracted through the fault report, and all faults of the same device can be analyzed.

[0059] Fault and fault causality: the fault caused by the failure of the device is extracted through the fault report, which is an important relationship for device fault correlation analysis and local to global influence analysis.

[0060] Fault phenomenon and fault cause relationship: the cause of the fault phenomenon is described, and the cause of the same fault phenomenon can be explored combined with the device to carry out device fault traceability analysis.

[0061] Fault and fault phenomenon relationship: the phenomenon when the fault occurs is extracted through the fault report, and the faults with the same phenomenon can be classified and viewed.

[0062] System and professional belonging relationship: describe the professional to which the system belongs, which is used to explore the subsystem of the professional.

[0063] Device and system belonging relationship: describe the subsystem to which the device belongs, which is used to explore the device under the subsystem.

[0064] Device and line belonging relationship: describe the line to which the device belongs, and explore the device according to the line.

[0065] Device and location location: describe the location information of the device, and view all devices at the location according to the location.

[0066] Device and supplier supply relationship: describe the supply manufacturer of the device, and explore the devices supplied by the same supplier.

[0067] Fault and fault report belonging relationship: describe the source report information of the fault.

[0068] Fault and inspection operation correlation relationship: the inspection operation recently done before the fault occurs is extracted through the fault report, and the device fault factor analysis can be carried out combined with the device.

[0069] Fault report and personnel filling relationship: describe the reporter of the fault report, and explore all fault reports of the same reporter.

[0070] Fault report and personnel audit relationship: describes the auditors of the fault report, which can explore all fault reports of the same auditor.

[0071] Personnel and professional affiliation: describes the professional information of the personnel, which can explore the personnel of the same professional.

[0072] Personnel and route affiliation: describes the route information of the personnel, which can explore the personnel according to the route.

[0073] Station and line affiliation: describes the line to which the station belongs, and all stations on the line can be explored through the line.

[0074] Interval and line relationship: describes the line on which the interval is located, and all intervals on the line can be explored through the line.

[0075] Main station and line relationship: describes the line on which the main station is located, and all main stations on the line can be explored through the line.

[0076] Depot and line relationship: describes the line on which the depot is located, and all depots on the line can be explored through the line.

[0077] S102, obtaining a fault report, analyzing the fault report to obtain analysis result data;

[0078] Specifically, the fault report is mainly used to reflect the equipment fault handling process, including fault description, fault cause analysis, fault solution measures, etc.

[0079] According to the analysis of the fault report, the fault report is filled in by using a unified template, and the filling content includes fault report name, filling department, filling time, filler, auditor, basic situation, handling situation, cause analysis, measures taken, follow-up measures, etc.

[0080] The data to be extracted in the fault report includes fault name, filling department, filling time, filler, auditor, fault phenomenon, fault cause, fault influencing factor, fault equipment, fault component, fault component problem, etc.

[0081] Among them, the fault name, filling department, filling time, filler, auditor are obtained in the heading information of the fault report according to the rules, and the fault phenomenon is obtained in the name of the fault report according to the rules.

[0082] Fault equipment, fault component, fault component problem are extracted in the conclusion section of the fault report by text extraction model. When using knowledge extraction model extraction, a large number of historical fault reports need to be provided as corpus for annotation before knowledge extraction model training.

[0083] The analysis result data includes an influence relationship between devices, an association relationship between devices and faults, and a causal relationship between faults and faults.

[0084] The fault report is corpus annotated, the knowledge extraction model is trained through corpus annotation, and a trained knowledge extraction model is obtained.

[0085] The trained knowledge extraction model is used to extract knowledge from the fault report to obtain extracted data.

[0086] In the scheme, the following steps can be used: the corpus annotation is divided into a training set, a validation set and a test set; the knowledge extraction model is trained based on the training set; the performance of the trained knowledge extraction model is evaluated based on the validation set to obtain a knowledge extraction model that meets the performance condition; the generation result of the knowledge extraction model that meets the performance condition is evaluated based on the test set to obtain an evaluation index corresponding to the knowledge extraction model;

[0087] The performance of the trained knowledge extraction model is evaluated based on the validation set to obtain a knowledge extraction model that meets the performance condition; the similarity calculation result of the knowledge extraction model that meets the performance condition is evaluated based on the test set to obtain an evaluation index corresponding to the knowledge extraction model. The performance of the knowledge extraction model is evaluated to obtain a percentage score (i.e. the highest score is 100 points and the lowest score is 0 points), and the knowledge extraction model with a score greater than a set value is determined based on the percentage score, for example, the knowledge extraction model with a score greater than 90 points is the knowledge extraction model that meets the performance condition;

[0088] The knowledge extraction model that meets the performance condition is evaluated to obtain an evaluation index of the knowledge extraction model, and the evaluation value corresponding to each evaluation index is calculated, which is used to represent the ability value of the knowledge extraction model on the evaluation index.

[0089] The interface between the front and back ends of the fault report analysis is used for report uploading, report list query, analysis result modification and saving, analysis result review, analysis result query and report original file query;

[0090] The interface between the back end and the knowledge graph platform is used for entity relationship extraction, entity relationship attribute data saving and entity data query.

[0091] S103, based on device maintenance related data and analysis result data, a device fault graph is constructed;

[0092] Specifically, as shown in the concept attribute description table in Table 1, the equipment fault graph includes entities, attributes, relationships, object attributes, and numerical attributes; the entity is a representation of a knowledge item in the knowledge graph, the attribute is a representation of other accessory information of the knowledge item in the knowledge graph, the relationship is a representation of the association between entities in the knowledge graph, the object attribute is a definition of the relationship by the knowledge graph platform, and the numerical attribute is a definition of the attribute by the knowledge graph platform.

[0093] Table 1

[0094] As shown in FIG. 2, the equipment fault graph is constructed to organize the relationships between equipment and faults, equipment and equipment, and faults and faults, thereby providing a basic data structure for equipment fault analysis and intelligent retrieval of equipment faults.

[0095] The creation, modification, and deletion operations of the concepts, relationships, and attributes are mainly implemented by the knowledge modeling module of the knowledge graph management platform.

[0096] The equipment fault graph includes 17 kinds of data such as specialty, system, equipment, supplier, location, line, station, section, main station, vehicle depot, personnel, fault report, fault, fault phenomenon, fault cause, and fault factor, and 21 kinds of relationships such as the parent relationship between equipment and equipment, the influence relationship between equipment and equipment, the association relationship between equipment and fault, the membership relationship between fault and fault report, and the cause-and-effect relationship between faults. The equipment fault graph constitutes a fault system through fault-related knowledge such as fault, fault equipment, fault resolution measure, fault phenomenon, fault cause, and fault influencing factor.

[0097] The LMIS system provides structured specialty, system, line, location, equipment, personnel, and supplier data. The fault report provides unstructured doc files, from which data such as fault occurrence time, fault name, fault phenomenon, fault cause, and fault equipment can be extracted. The structured and unstructured data are accessed and processed, and finally written to the equipment fault graph for storage.

[0098] The extracted data and the equipment maintenance related data are fused to obtain a data fusion result; and the equipment fault graph is adjusted based on the data fusion result.

[0099] The fusion includes table structure fusion and data fusion.

[0100] The table structure fusion fuses the fault structure in the work order in the LMIS system and the fault structure in the fault report to form a fault system constructed by the knowledge of fault, fault phenomenon, fault cause, fault solution measure, fault equipment, fault component, fault influencing factor and fault influence surface.

[0101] The data fusion includes equipment, fault, fault phenomenon and fault cause. The equipment, fault name, fault phenomenon and fault cause obtained from the fault report only have text description or name, so the data fusion takes the data in the LMIS system as the main data, and if there is the same entity, the attribute information in the LMIS system is retained. The data fusion content is described as follows,

[0102] Equipment: The equipment in the LMIS system and the equipment name parsed from the fault report are fused as entities, and the attribute information in the LMIS system is retained during the fusion.

[0103] Fault: The fault information in the graph and the fault name parsed from the fault report are fused as entities, and if there is, the attribute in the LMIS system is retained, and if not, it is added.

[0104] Fault phenomenon: The fault phenomenon in the graph and the fault phenomenon parsed from the fault report are fused as entities, and if there is, the attribute in the LMIS system is retained, and if not, it is added.

[0105] Fault cause: The fault cause in the graph and the fault cause parsed from the fault report are fused as entities, and if there is, the attribute in the LMIS system is retained, and if not, it is added.

[0106] The vector values of all equipment names are calculated, and the vector values are saved to the vector library;

[0107] The fault report is parsed to obtain the equipment name, and the vector calculation is performed on the equipment name to obtain the equipment name vector value;

[0108] The equipment entity name with the similarity in the threshold range is queried from the vector library through the equipment name vector value.

[0109] In machine learning and natural language processing, vectorization (embedding) refers to the process of mapping high-dimensional data (such as text, pictures, audio) to a low-dimensional space. The vectorization result is usually a vector composed of real numbers, which represents the input data as a point in a continuous numerical space. In simple terms, the vectorization result is an N-dimensional real value vector, which can be used to represent almost anything, such as text, music, video, etc.

[0110] The vectorization results of real-valued vectors can represent the semantics of words, mainly because these vectorization results are learned according to the occurrence patterns of words in linguistic contexts. For example, if a word often appears with another word in some contexts, then the vectorization results of these two words will have similar positions in the vector space, which means that they have similar meanings and semantics.

[0111] The semantics of words can be represented by their distribution in context, that is, the meaning of a word can be inferred from its surrounding words. Large language models such as BERT, ELMo, and GPT can generate context-dependent vectorization representations that can better capture the semantics and context information of words.

[0112] Determine whether the device name and the device entity name are correct. If the device name and the device entity name are not correct, reselect.

[0113] The interface between the front and back ends of the device fault analysis includes: system, device, fault retrieval interface, knowledge card query interface, device fault association analysis interface, local to global impact analysis interface, fault influencing factor analysis interface, fault traceability analysis interface, according to fault query related device interface, according to device query related fault interface, according to system query related fault interface and according to fault query related system interface;

[0114] S104, integrate the device fault atlas into the knowledge graph platform.

[0115] Specifically, the knowledge graph platform includes a search home page, a search result module, and a knowledge exploration module.

[0116] The search home page includes full-text retrieval, classification drop-down box, search history record, advanced search, intent recognition, search association prompt, search hot list, popular data, and high-frequency case recommendation functions.

[0117] Full-text retrieval: provides a unified search entry and supports unified full-text retrieval function; according to the user input search term, provides unified or dimension-based full-text retrieval capability. Support search filtering, display retrieval results according to filtering dimensions. Support line, professional, system, device, fault component, fault type, fault, fault phenomenon, fault reason, and fault influencing factor data retrieval.

[0118] Classification drop-down box: provides a classification retrieval entry and supports data retrieval by classification;

[0119] Search history: display the user's search history keywords, show the last 5; according to the current user behavior analysis, the accumulation of the current user's search history, display the last 5 keywords, support click history search into search details.

[0120] Advanced search: describe the type of device belongs to, can explore all the devices of this type; Support "accurate" "fuzzy" selection way of advanced search, you can choose search field by keyword "accurate" or "fuzzy" search, support multi-condition joint search.

[0121] Intention recognition: describe the parent device of the device, can be constructed according to the relationship of the device tree combined with the device fault to realize the influence analysis from local to whole; Support semantic intention recognition based on graph, user input search statement system according to the concept, entity, attribute, relationship in the graph mode to analyze and identify user intention, return the corresponding result.

[0122] Search association prompt: according to the user input key words for association prompt; Support input key words association prompt, including system, professional, equipment, fault phenomenon, fault reason and other types of data title association prompt.

[0123] Search hot list: show all the top ten keywords of user search frequency; According to all user behavior analysis and statistics, accumulate user search records, recommend and display high frequency search words, show the top ten search hot words, support "change a change" update search hot list, support click history search into search details.

[0124] Hot data: display all the top ten knowledge entries clicked by users; According to user behavior analysis and statistics, accumulate user high frequency click knowledge in search list, recommend and display high frequency knowledge, display the top ten hot data, support click knowledge entry into knowledge exploration page.

[0125] High frequency case recommendation: display the top ten fault phenomena of fault occurrence frequency. According to the analysis and statistics of the current system fault knowledge, accumulate the same fault knowledge frequency, recommend and display the high frequency fault knowledge according to the occurrence frequency, display the top ten fault knowledge, support "change a change" update search case, support click high frequency case into knowledge exploration page.

[0126] The search result module includes result list, data classification statistics, search result sorting, knowledge cloud and associated recommendation;

[0127] Result list: display the search result list;

[0128] According to the user input search word or other search way, such as advanced search, history recommendation, etc., enter the search result page.

[0129] Different types of knowledge display fields are as follows:

[0130] Line data: display knowledge title, type label, data creation time, line name, operating mileage, number of stations, line opening date.

[0131] Professional data: display knowledge title, type label, data creation time, professional name.

[0132] System data: display knowledge title, type label, data creation time, system name.

[0133] Equipment data (fault components): display knowledge title, type label, data creation time, asset code, supplier, manufacturer, usage status, cross-professional fault correlation analysis, local-to-global impact analysis, fault occurrence factor correlation analysis, equipment fault cause tracing analysis.

[0134] Fault type: display knowledge title, type label, data creation time, type name.

[0135] Fault phenomenon data: display knowledge title, type label, data creation time, fault phenomenon, fault cause, solution measures, cross-professional fault correlation analysis, local-to-global impact analysis, fault occurrence factor correlation analysis, equipment fault cause tracing analysis.

[0136] Fault data: display knowledge title, type label, data creation time, fault name, fault phenomenon, fault cause, solution measures, cross-professional fault correlation analysis, local-to-global impact analysis, fault occurrence factor correlation analysis, equipment fault cause tracing analysis.

[0137] Fault cause data: display knowledge title, type label, data creation time, fault cause, cause description.

[0138] Fault factor data: display knowledge title, type label, data creation time, name, factor type, description.

[0139] Data classification statistics; classified statistics according to search results; according to the classification aggregation items configured by the background, the search list can be aggregated by the knowledge filtering items.

[0140] Search result sorting: support sorting operation on the result list; support basic sorting according to publishing time and knowledge relevance; support business-based mixed sorting, which allows users to select sorting fields, and the sorting fields support multiple selection;

[0141] The sortable fields for each type of data are as follows:

[0142] Device: title, date of use.

[0143] Fault: title, date of occurrence.

[0144] Other: title, date of publication.

[0145] Knowledge cloud map: display the identified entity triple information; support entity positioning based on user input search words, associate knowledge graph, and display relevant knowledge graph fragment information based on the positioned entity.

[0146] Association recommendation: recommend related devices and related faults. Based on user search data, make associated recommendations, and recommend related devices and related fault phenomena based on the association of hit entities. Search system, specialty, line, location, etc. make associated device recommendations based on their associated devices, and make fault phenomenon recommendations based on the associated faults of the associated devices. Search for devices, recommend devices based on the impact relationship of the devices, and recommend fault phenomena based on the associated faults of the devices. Search for faults, make associated device recommendations based on the associated devices of the faults, and recommend fault phenomena based on the cause-and-effect relationship of the faults.

[0147] The knowledge exploration module includes entity search, precise search, secondary screening, entity paging, entity positioning, graph layout switching, re-layout, node personalization settings, and knowledge cards.

[0148] Entity search: fuzzy search on entities, default is prefix matching, can be modified to full-text matching.

[0149] The search bar in the upper left corner of the knowledge application page is used to search for entities (supports searching for categories in specific pages), and the system switches the center (i.e. focus) of visual exploration and analysis based on the target entity of the search, and displays the network relationship graph centered on the target entity.

[0150] Entity search supports prefix search and fuzzy search (i.e. full-text search) two search methods. The system uses pattern search by default. The statements of keyword matching entities, keyword search, entity search, etc. are all based on prefix search or fuzzy search methods.

[0151] 1) Prefix search uses keywords to match entities from the beginning, returning entities that meet the conditions; for example, the keyword "China" can match the entity "People's Republic of China", but cannot match the entity "Love My China".

[0152] 2) Fuzzy search is based on the automatic word segmentation results of entities and search keywords for matching. For example, the keyword "Chinese" can match the entity "Chinese community" (segmentation result: Chinese / community), but cannot match the entity "Chinese people" (segmentation result: China / people).

[0153] Exact search: Provide multi-dimensional knowledge search of name, attribute, relationship; Exact search supports entity search using keyword, category, attribute constraint. Exact search can be entered by inputting space in the search bar and clicking the "Exact search" on the right side, or by selecting the first line of keywords in the drop-down prompt when entering keywords.

[0154] Secondary screening: Secondary filtering of canvas display results; Secondary screening is used to further filter the network relationship diagram displayed on the current page, generating a subgraph that meets the constraint conditions. Secondary screening conditions include subject attribute screening and relationship attribute screening. Subject attribute screening uses the numerical attribute of the entity as a constraint to filter entities. Relationship attribute screening uses the edge numerical attribute of the relationship as a constraint to filter relationships.

[0155] Entity pagination: Display nodes on the canvas by pagination; The number of entities displayed per page and page turning: The number of entities that can be directly associated with the focus entity can be set. When the number of entities associated with the focus is greater than the number of entities displayed per page, use page turning to view entities.

[0156] Entity positioning: Positioning on the canvas according to entity name; When there are many entities displayed on the canvas and the target entity needs to be quickly found, entity positioning can be used. Click the icon, enter the entity keyword in the pop-up window, select the target entity, and the system will highlight the target entity.

[0157] Map layout switching: Switch the layout of the canvas to support tree, dynamic layout, static layout, etc.; The system provides multiple map visualization layouts, including: dynamic layout, tree layout, reverse tree layout, left tree layout, right tree layout, radial layout, radial force layout, Sanji layout, rectangular tree map layout, sunrise layout, vertical layout, horizontal layout, ring layout, package layout, static layout. The appropriate map layout for visualization analysis and exploration can be selected by clicking the icon and selecting the appropriate map layout.

[0158] Relayout: Initialize the layout of the canvas; After dragging the map in the canvas, click the icon, and the system will rearrange the entities and relationships of the map; The functions provided at the bottom right of the knowledge application map canvas are: back, export, full screen / only main view display, drag selection, highlight hover element, thumbnail, zoom in / zoom out.

[0159] Node personalization settings; Support for personalizing node display; In the knowledge application map canvas, click the right mouse button when the mouse is over an entity to provide related entity operation options: layout, switch focus, node exploration, map jump, hide node, add subgraph, entity style, other operations, clear, general settings.

[0160] Layout: Layout can be set for one or a group of entity nodes. Dragging and selecting can be used to set the layout for a group of entity nodes.

[0161] Switch focus: If the currently selected entity is not the focus entity, the "Switch focus" function can be used to switch the entity to the focus of the graph exploration.

[0162] Node exploration: If the currently selected entity is not the focus entity and you want to explore the relationship of associated entities, you can use the "Node exploration" function. When the number of nodes associated with the entity is greater than the number of entities per page, you can right-click to view the next page.

[0163] Graph jump: Provides a jump to the graph exploration, time series graph exploration, and graph analysis Pro page with the selected entity as the focus.

[0164] Hide node: During graph exploration analysis, if you want to hide some entity nodes, you can right-click to select "Hide node". Please note that after hiding the node, you cannot re-display the entity node through the "Back" operation.

[0165] Add subgraph: You can add a subgraph based on the current analysis results. Right-click to select "Add subgraph", enter the entity to be explored in the pop-up window, add it to the analysis and confirm, and the system will return to the new subgraph.

[0166] Entity style: Set the style of a single entity, including: node virtualization, node highlighting, label built-in, hidden label, style configuration (color, label, label color, radius, display form, text embedding, picture, cropping).

[0167] Other operations: Provides the function of locking nodes.

[0168] Clear: Clear the canvas;

[0169] General settings: Configure the styles of the canvas background, node highlighting / virtualization, and edge highlighting / virtualization. Please note that this configuration is not permanent and will be restored to the default value after refreshing the page.

[0170] Knowledge card: Show the attribute information of the node; the "Details" function bar on the right side of the knowledge application page displays the knowledge card information of the entity, including the entity's picture, start time-end time, confidence level, type (category), source, introduction, and numerical attributes (grouped by attributes).

[0171] The "File" function bar on the right side of the knowledge application page displays the multi-modal data information of the entity, including the entity's related pictures, documents, videos, and audio.

[0172] The multi-modal data of the entity can be added in the entity editing page of the "graph management-knowledge editing". Or upload the file in the "file storage", and then perform the entity and the multi-modal data through the "online indexing".

[0173] The "filter" function bar on the right side of the knowledge application page configures the filtering conditions for the graph exploration analysis. The configuration information includes:

[0174] 1) Set the number of display layers: the number of layers from the focus to the outermost entity;

[0175] 2) Set the analysis subject: the category range of the exploration;

[0176] 3) Set the analysis relationship: the relationship range of the exploration;

[0177] 4) Set the expansion level: the number of layers when exploring non-focus entities;

[0178] 5) Set the direction of the query relationship: bidirectional / forward / backward;

[0179] 6) Relationship display merging: merge / do not merge, merge refers to that if there are multiple relationships between two entities, only one connection line is displayed in the canvas;

[0180] 7) Set the confidence range of the query node and relationship: the range is 0-1;

[0181] 8) Set the weight range of the query node and relationship: the range is 0-1;

[0182] 9) Set the number of high-level nodes: the range is 10-50, which is used to expand to the third layer or higher layer outside the focus entity, and each high-level contains a total number of nodes;

[0183] Set the attribute value condition: including subject attribute filtering and relationship attribute filtering.

[0184] S105, receiving the search word input by the current user in the search box, obtaining the search result corresponding to the search word, and displaying the search result on the search result page;

[0185] Specifically, the search result includes a result list, data classification statistics, search result sorting, a knowledge cloud diagram, and associated recommendations.

[0186] The device fault graph is used to perform entity search on the search word input by the current user, and a target entity is obtained;

[0187] The number of pages of the associated entity corresponding to the target entity is set in advance, and it is judged whether the number of associated entities is greater than the number of pages of the associated entity corresponding to the target entity. If yes, the entity is paginated;

[0188] The entity search includes a prefix search, a fuzzy search, and an accurate search, and the accurate search includes entity search by keyword and / or category and / or attribute constraint;

[0189] The center of the knowledge cloud map is switched based on the target entity, and a knowledge cloud map centered on the target entity is displayed.

[0190] The knowledge cloud map is filtered again through secondary screening to generate a subgraph that meets the constraint condition, wherein the secondary screening condition includes subject attribute screening and relationship attribute screening.

[0191] S106, providing an operation entrance of knowledge exploration in the search result page, and the user views the related knowledge exploration of the current knowledge by triggering the operation entrance of knowledge exploration;

[0192] Specifically, the devices and faults associated with the to-be-analyzed starting point are sequentially connected in an influence relationship through the device fault atlas, and a device fault association influence path is obtained.

[0193] The path influence probability is calculated according to the path relationship quantity and the fault occurrence frequency.

[0194] Based on the device fault association influence path and the path influence probability, a device fault analysis result is obtained, wherein the to-be-analyzed starting point includes a starting point device and a starting point fault.

[0195] Based on the starting point device, a device entity having an influence relationship with the starting point device is inferred in the device fault atlas.

[0196] Based on the device entity, an associated fault corresponding to the device entity is inferred.

[0197] Based on the starting point fault, a cause-effect relationship fault corresponding to the starting point fault is inferred.

[0198] The cause-effect relationship fault and the associated fault are intersected to obtain reserved data, and the reserved data is taken as a path node.

[0199] Based on the path node, continuous inference is performed until the path node is empty.

[0200] The device fault atlas uses path analysis, atlas association analysis, and inference analysis to perform multi-dimensional calculation on the fault atlas, provides cross-professional device fault association analysis, device local-to-global influence analysis, fault factor association analysis, and other fault analysis functions, and provides device fault association traceability and other fault analysis functions.

[0201] The vehicle professional device fault unified retrieval method based on the knowledge graph further includes:

[0202] In response to a user-initiated knowledge card viewing request, a knowledge card corresponding to the target entity is displayed.

[0203] A fault knowledge index is constructed, wherein the fault knowledge index includes a classification drop-down box, search history records, device types, intent recognition, search association prompts, search hot lists, popular data, and high-frequency case recommendations.

[0204] Device information and fault information are searched based on the fault knowledge index.

[0205] The vehicle professional equipment fault unified retrieval method based on the knowledge graph acquires equipment repair related data, takes the equipment as the core, establishes an association relationship between the equipment repair related data, wherein the equipment repair related data includes asset data, location data, personnel data, work order data, fault code data, professional information data, and equipment subsystem code data; acquires a fault report, analyzes the fault report to obtain analysis result data, wherein the analysis result data includes an influence relationship between equipment and equipment, an association relationship between equipment and faults, and a cause-and-effect relationship between faults and faults; constructs an equipment fault graph based on the equipment repair related data and the analysis result data; integrates the equipment fault graph into a knowledge graph platform; receives a search word input by a current user in a search box, acquires a search result corresponding to the search word, and displays the search result on a search result page, wherein the search result includes a result list, data classification statistics, search result sorting, a knowledge cloud map, and association recommendations; provides an operation entry for knowledge exploration in the search result page, and a user views related knowledge exploration of current knowledge by triggering the operation entry for knowledge exploration. The problem of inaccurate retrieval results in the prior art is solved.

[0206] FIG. 3 is a flowchart of an embodiment of a vehicle professional equipment fault unified retrieval system based on the knowledge graph; as shown in FIG. 3, the vehicle professional equipment fault unified retrieval system based on the knowledge graph provided in the embodiment includes the following modules:

[0207] The first obtaining module 10 is configured to obtain equipment maintenance related data, establish an association relationship between the equipment maintenance related data taking equipment as a core, and the equipment maintenance related data comprises asset data, location data, personnel data, work order data, fault code data, professional information data and equipment subsystem code data; the association relationship comprises a parent-child relationship between equipment and equipment, an influence relationship between equipment and equipment, an association relationship between a fault and equipment, a cause-effect relationship between a fault and a fault, a cause-effect relationship between a fault phenomenon and a fault cause, a relationship between a fault and a fault phenomenon, an affiliation relationship between a system and a profession, an affiliation relationship between equipment and a system, an affiliation relationship between equipment and a line, a location of equipment and a location, a supply relationship between equipment and a supplier, an association relationship between a fault and an inspection operation, a reporting relationship between a fault report and personnel, an auditing relationship between a fault report and personnel, an affiliation relationship between personnel and a profession, an affiliation relationship between personnel and a line, an affiliation relationship between a station and a line, a relationship between a section and a line, a relationship between a main depot and a line, and a relationship between a vehicle depot and a line.

[0208] The second obtaining module 20 is configured to obtain a fault report, analyze the fault report to obtain analysis result data, and the analysis result data comprises an influence relationship between equipment and equipment, an association relationship between equipment and a fault, and a cause-effect relationship between a fault and a fault.

[0209] The fault report is subjected to corpus annotation, a knowledge extraction model is trained through corpus annotation, a trained knowledge extraction model is obtained, the trained knowledge extraction model is used for knowledge extraction on the fault report, extraction data is obtained, the extraction data is subjected to data fusion with the equipment maintenance related data, data fusion result is obtained, and the equipment fault graph is adjusted based on the data fusion result.

[0210] The graph construction module 30 is configured to construct an equipment fault graph based on the equipment maintenance related data and the analysis result data. Vector values of all equipment names are calculated, and the vector values are saved to a vector library; the fault report is analyzed to obtain equipment names, vector calculation is performed on the equipment names to obtain equipment name vector values; equipment entity names with a similarity within a threshold range are queried from the vector library through the equipment name vector values. It is judged whether the equipment names and the equipment entity names are correct, and if the equipment names and the equipment entity names are not correct, reselection is performed.

[0211] The integration module 40 is configured to integrate the equipment fault graph into a knowledge graph platform.

[0212] The search module 50 is configured to receive a search word input by a current user in a search box, acquire a search result corresponding to the search word, and display the search result on a search result page, wherein the search result comprises a result list, data classification statistics, search result sorting, a knowledge cloud diagram, and associated recommendations.

[0213] The exploration module 60 is configured to provide an operation entry of knowledge exploration in the search result page, and a user views relevant knowledge exploration of current knowledge by triggering the operation entry of knowledge exploration.

[0214] FIG. 4 is a schematic diagram of an electronic device according to an embodiment of the present application. As shown in FIG. 4, the electronic device 70 comprises a processor 701, a memory 702, and a bus 703.

[0215] The processor 701 and the memory 702 are connected to each other through the bus 703.

[0216] The processor 701 is configured to call program instructions in the memory 702 to execute the method provided by each method embodiment described above, for example, comprising: acquiring device maintenance related data, taking a device as a core, establishing an association relationship between the device maintenance related data, wherein the device maintenance related data comprises asset data, location data, personnel data, work order data, fault code data, professional information data, and device subsystem code data; acquiring a fault report, analyzing the fault report to obtain analysis result data, wherein the analysis result data comprises an influence relationship between devices, an association relationship between a device and a fault, and a cause-effect relationship between faults; constructing a device fault graph based on the device maintenance related data and the analysis result data; integrating the device fault graph into a knowledge graph platform; receiving a search word input by a current user in a search box, acquiring a search result corresponding to the search word, and displaying the search result on a search result page, wherein the search result comprises a result list, data classification statistics, search result sorting, a knowledge cloud diagram, and associated recommendations; providing an operation entry of knowledge exploration in the search result page, and a user views relevant knowledge exploration of current knowledge by triggering the operation entry of knowledge exploration.

[0217] The embodiment provides a non-transitory computer readable storage medium, which stores computer instructions, and the computer instructions cause a computer to execute the method provided by each method embodiment, for example, comprising the following steps: obtaining equipment maintenance related data, taking equipment as a core, and establishing an association relationship between the equipment maintenance related data, wherein the equipment maintenance related data comprises asset data, position data, personnel data, work order data, fault code data, professional information data and equipment subsystem code data; obtaining a fault report, analyzing the fault report to obtain analysis result data, wherein the analysis result data comprises an influence relationship between equipment and equipment, an association relationship between equipment and a fault, and a cause-effect relationship between a fault and a fault; constructing an equipment fault graph based on the equipment maintenance related data and the analysis result data; integrating the equipment fault graph into a knowledge graph platform; receiving a search word input by a current user in a search box, obtaining a search result corresponding to the search word, and displaying the search result on a search result page, wherein the search result comprises a result list, data classification statistics, search result sorting, a knowledge cloud map and associated recommendation; and providing an operation entrance of knowledge exploration in the search result page, and the user views related knowledge exploration of current knowledge by triggering the operation entrance of knowledge exploration.

[0218] The apparatus embodiments described above are merely illustrative, wherein the units shown as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0219] Although the present application has been described in detail above with general description and specific embodiments, some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, these modifications or improvements made on the basis of not deviating from the spirit of the present application, are within the scope of the present application.

Claims

The application discloses a knowledge graph-based unified retrieval method for vehicle professional equipment faults. The method comprises: acquiring equipment maintenance related data, taking equipment as the core, establishing the association relationship between the equipment maintenance related data, wherein the equipment maintenance related data comprises asset data, location data, personnel data, work order data, fault code data, professional information data and equipment subsystem code data; acquiring a fault report, analyzing the fault report to obtain analysis result data, wherein the analysis result data comprises the influence relationship between equipment and equipment, the association relationship between equipment and fault, and the causal relationship between faults and faults; constructing an equipment fault graph based on the equipment maintenance related data and the analysis result data; integrating the equipment fault graph into a knowledge graph platform; receiving a search term input by a current user in a search box, acquiring a search result corresponding to the search term, and displaying the search result on a search result page, wherein the search result comprises a result list, data classification statistics, search result sorting, a knowledge cloud map and associated recommendations; providing a knowledge exploration operation entry in the search result page, and allowing the user to view related knowledge exploration of current knowledge by triggering the knowledge exploration operation entry. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 1 is characterized in that, The association relationship comprises a parent-child relationship between equipment and equipment, an influence relationship between equipment and equipment, an association relationship between fault and equipment, a causal relationship between faults and faults, a cause relationship between fault phenomenon and fault cause, a relationship between fault and fault phenomenon, a belonging relationship between system and profession, a belonging relationship between equipment and system, a belonging relationship between equipment and line, a location of equipment and location, a supply relationship between equipment and supplier, a belonging relationship between fault and fault report, an association relationship between fault and inspection operation, a reporting relationship between fault report and personnel, an auditing relationship between fault report and personnel, a belonging relationship between personnel and profession, a belonging relationship between personnel and route, a belonging relationship between station and line, a relationship between section and line, a relationship between main station and line, and a relationship between vehicle depot and line. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 1 is characterized in that, The acquiring of the fault report, the analyzing of the fault report, and the obtaining of the analysis result data comprise: performing corpus annotation on the fault report, training a knowledge extraction model through corpus annotation to obtain a trained knowledge extraction model; performing knowledge extraction on the fault report through the trained knowledge extraction model to obtain extraction data; performing data fusion on the extraction data and the equipment maintenance related data to obtain data fusion results; adjusting the equipment fault graph based on the data fusion results. The construction of the equipment fault graph based on the equipment maintenance related data and the analysis result data comprises: The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 1 is characterized in that, calculating vector values of all equipment names and saving the vector values to a vector library; analyzing the fault report to obtain equipment names, and performing vector calculation on the equipment names to obtain equipment name vector values; querying equipment entity names with a similarity within a threshold range through the equipment name vector values to the vector library; determining whether the equipment names and the equipment entity names are correct, and if the equipment names and the equipment entity names are not correct, reselecting. ​ The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 1 is characterized in that, The receiving of the search term input by the current user in the search box, the obtaining of the search result corresponding to the search term, and the display of the search result on the search result page, comprising: The target entity is obtained by performing entity search on the search term input by the current user through the device fault graph; The entity search includes prefix search, fuzzy search and accurate retrieval, and the accurate retrieval includes entity search through keyword and / or category and / or attribute constraint; The center of the knowledge cloud map is switched based on the target entity, and the knowledge cloud map centered on the target entity is displayed. The vehicle professional device fault unified retrieval method based on the knowledge graph according to claim 5 The method is characterized in that, The target entity is obtained by performing entity search on the search term input by the current user through the device fault graph, comprising: The number of associated entities corresponding to the target entity is set as a per-page display threshold, and it is judged whether the number of associated entities is greater than the per-page display threshold, and if so, the entity is paginated. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 5 is characterized in that, The receiving of the search term input by the current user in the search box, the obtaining of the search result corresponding to the search term, and the display of the search result on the search result page, further comprising: The search term is associated with the recommendation, and the related device recommendation and the related fault phenomenon recommendation are performed based on the association relationship of the target entity. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 5 is characterized in that, The receiving of the search term input by the current user in the search box, the obtaining of the search result corresponding to the search term, and the display of the search result on the search result page, further comprising: The knowledge cloud map is filtered again through secondary screening to generate a subgraph that meets the constraint condition, wherein the secondary screening condition includes subject attribute screening and relationship attribute screening. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 1 is characterized in that, The vehicle professional device fault unified retrieval method based on the knowledge graph, further comprising: A fault knowledge index is constructed, wherein the fault knowledge index includes a classification drop-down box, search history records, device types, intent recognition, search association prompts, search hot lists, popular data and high-frequency case recommendations; Device information and fault information are searched based on the fault knowledge index. The knowledge graph-based vehicle professional equipment fault unified retrieval method according to claim 7, characterized in that, The vehicle professional device fault unified retrieval method based on the knowledge graph, further comprising: In response to a knowledge card viewing request initiated by the user, a knowledge card corresponding to the target entity is displayed.

Images