A knowledge graph-based work order whole-process electronic management method and system

Abstract

The application discloses a kind of based on knowledge graph's work order whole process electronization management method and system, it is related to work order management technical field, including: definition work order management field core entity, entity interrelation to obtain core entity relationship chain, determine each entity attribute and construct field ontology diagram;Structured and unstructured data are divided and respectively preprocessed, and entity-relation-attribute chain is obtained;Accordingly, build initial knowledge graph, after its service is connected with work order management system;Integrating work order initiation channel using system, data acquisition, data are classified according to initial graph entity-relation and are matched with optimal resource, and the standard graph of optimization graph is obtained;Optimized solution and knowledge push list are generated in combination with standard graph, and work order processing process information is recycled;Archive work order and secondary optimization graph obtain preferred standard graph.The advantage lies in: realize work order whole process intelligentization, improve processing efficiency, guarantee knowledge sediment reuse.

Description

Technical Field

[0001] This invention relates to the field of work order management technology, specifically to a knowledge graph-based method and system for full-process electronic management of work orders. Background Technology

[0002] Work order management is a core component of enterprise operations, broadly covering various scenarios such as IT maintenance, customer service, and equipment repair. Its processing efficiency and knowledge accumulation capabilities directly impact service quality and operating costs. As business complexity increases, the volume and types of work orders surge. Traditional manual allocation, paper records, or simple electronic systems suffer from issues such as non-standardized processes, fragmented information, and difficulty in knowledge reuse. Knowledge graphs, with their semantic association capabilities, provide support for intelligent management of the entire work order process. Therefore, developing a knowledge graph-based electronic management method and system for the entire work order process is of significant practical importance for improving work order processing efficiency and achieving closed-loop knowledge accumulation.

[0003] Existing knowledge graph-based methods and systems for full-process electronic management of work orders fail to effectively balance structured and unstructured data during data processing, which can easily lead to missing implicit information or low processing efficiency. Furthermore, existing technologies rely on human experience for ontology construction, and the definition of core entities, relationships, and attributes lacks standardized processes, resulting in poor adaptability of ontology graphs. Therefore, it is necessary to provide a knowledge graph-based method and system for full-process electronic management of work orders to address the aforementioned problems. Summary of the Invention

[0004] To address the aforementioned technical issues, this paper provides a knowledge graph-based electronic management method and system for the entire process of work orders. This technical solution solves the problems mentioned in the background section regarding existing knowledge graph-based electronic management methods and systems for the entire process of work orders. These methods and systems fail to effectively balance structured and unstructured data during data processing, which can easily lead to the loss of implicit information or low processing efficiency. Furthermore, existing technologies rely on human experience for ontology construction, and the lack of standardized processes for defining core entities, relationships, and attributes results in poor adaptability of ontology graphs.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A knowledge graph-based method for full-process electronic management of work orders includes:

[0007] Define the core entities in the work order management domain, thereby defining the relationships between the core entities and obtaining the core entity relationship chain. Then, determine the attributes corresponding to each core entity and construct the work order management domain ontology graph.

[0008] Based on the ontology graph of the work order management domain, work order data sources are collected and simultaneously divided into structured and unstructured data sources to obtain structured work order data and unstructured work order data. Then, data preprocessing is performed on the structured work order data and unstructured work order data respectively to obtain entity-relationship-attribute chain.

[0009] Based on the entity-relationship-attribute chain, an initial knowledge graph is constructed, and then the initial knowledge graph is serviced and integrated with the work order management system.

[0010] The work order management system is used to integrate work order initiation channels and collect electronic work order source data synchronously in real time. Then, the electronic work order source data is classified according to the entity-relationship in the initial knowledge graph, and the classified electronic work order source data is optimally matched with resources to obtain standard electronic work order source data. This data is used to perform the first optimization of the initial knowledge graph to obtain the standard knowledge graph.

[0011] Based on the standard knowledge graph and combined with real-time collected electronic work order source data, a list of preferred solutions and a knowledge push list are dynamically generated, and then work order processing flow information is collected.

[0012] Based on the work order processing flow information, the work orders are archived, and the standard knowledge graph is optimized a second time to obtain the preferred standard knowledge graph. Then, the work order management system is connected with the preferred standard knowledge graph.

[0013] In an optional embodiment, the process of defining core entities in the work order management domain, thereby defining the relationships between these core entities to obtain a core entity relationship chain, and then determining the attributes corresponding to each core entity to construct an ontology graph for the work order management domain, specifically includes:

[0014] Based on the business scenario requirements of the work order management field, we conducted domain research to obtain business process documents, historical work order records, and domain expert experience data related to work order management.

[0015] By using business process documents, historical work order records, and domain expert experience data, a candidate set of entities whose frequency of occurrence in historical work order records is greater than or equal to a preset threshold is selected.

[0016] Based on the candidate entity set, domain experts are invited to score the coreness of the candidate entities, and entities with a coreness score of 3 or higher are selected to obtain the core entity list of the work order management domain.

[0017] Analyze the interaction scenarios of each core entity in the business process, count the frequency of associations between entities, and determine the relationship weights between entities;

[0018] Filter out entity relationships with relationship weights greater than or equal to a preset relationship weight threshold, sort them from high to low according to association frequency, and obtain the core entity relationship chain;

[0019] Based on the core entity relationship chain and the business scenario requirements of the core entities, the attribute items corresponding to each core entity are determined, and the integrity requirement coefficient of each attribute item is determined. Attribute items with an integrity requirement coefficient greater than or equal to 0.9 are selected as core attributes.

[0020] Based on the core entity list, core entity relationship chain, and core entity core attributes, semantic constraints of entities, relationships, and attributes are defined to generate an ontology graph for the work order management domain.

[0021] In an optional embodiment, based on the ontology graph of the work order management domain, work order data sources are collected, and the work order data sources are simultaneously divided into structured and unstructured data to obtain structured work order data and unstructured work order data. Then, data preprocessing is performed on the structured and unstructured work order data respectively to obtain the entity-relationship-attribute chain, specifically including:

[0022] Based on the ontology diagram of the work order management domain, the collection scope of work order data sources is determined. The collection scope includes scanned copies of historical paper work orders, historical electronic work orders in the work order management system, business data from business-related systems, and processing manuals and training documents written by domain experts, forming the original work order data source set.

[0023] Based on the original work order data source set, the data format characteristics of each data source are extracted, and then the matching degree between the data source and the core fields of the ontology graph is determined.

[0024] Based on matching degree and data format characteristics, the original work order data source set is divided into work order structured data and work order unstructured data;

[0025] Redundant and abnormal data in the work order structure data are removed to obtain cleaned structured data;

[0026] Based on the cleaned structured data, data standardization is performed according to the attribute definition of the ontology graph, unifying the same attribute value in different formats into a preset format to obtain standardized structured data;

[0027] Text processing is performed on unstructured work order data to obtain text-based unstructured data.

[0028] The textual unstructured data is matched with the core entities in the ontology graph of the work order management domain to obtain entity matching results;

[0029] Based on the entity matching results, combined with the core entity relationship chain in the ontology graph of the work order management domain, the entity-relationship pairs in the unstructured data are determined, and the attribute information corresponding to the entities is extracted to obtain the entity-relationship-attribute set of the unstructured data.

[0030] Based on standardized structured data, the core entities, relations, and attributes in the ontology graph are directly mapped to obtain the entity-relationship-attribute set of structured data;

[0031] Data fusion is performed on the entity-relationship-attribute set of structured data and the entity-relationship-attribute set of unstructured data to obtain the entity-relationship-attribute chain.

[0032] In an optional embodiment, the step of constructing an initial knowledge graph based on the entity-relationship-attribute chain, and then service-izing the initial knowledge graph and interfacing it with the work order management system, specifically includes:

[0033] Based on the entity-relationship-attribute chain, determine the number of entities, relations and attributes in the chain, and thus obtain the scale of graph storage requirements.

[0034] Based on the storage requirements, determine the appropriate graph database, then create a knowledge graph template, define entity tags, relation types and attribute keys, and ensure consistency with the entity-relationship-attribute chain structure;

[0035] Based on the created knowledge graph template, data in the entity-relationship-attribute chain is imported into the graph database in batches;

[0036] After all the data has been imported, entity indexes and relationship indexes are built. The index fields include entity ID and core attributes. The query response time after the indexes are built is tested to determine the initial knowledge graph.

[0037] Develop a graph service API based on the initial knowledge graph;

[0038] Based on the knowledge graph service API, write API test cases and execute the test cases to complete the service-oriented transformation of the initial knowledge graph;

[0039] Obtain the interface documentation of the work order management system and analyze the system's interface types, request protocols, and data interaction formats;

[0040] Based on the interface information of the work order management system, an adapter for the graph service and the work order management system was developed to realize data format conversion and protocol adaptation between the two.

[0041] After the integration is completed, joint debugging tests are conducted to simulate the scenario of the work order management system calling the knowledge graph service API, thus completing the initial integration between the knowledge graph and the work order management system.

[0042] In an optional embodiment, the step of integrating work order initiation channels using the work order management system, synchronously collecting electronic work order source data in real time, then classifying the electronic work order source data according to the entity-relationship in the initial knowledge graph, and performing optimal resource matching on the classified electronic work order source data to obtain standard electronic work order source data, which is used to perform the first optimization of the initial knowledge graph to obtain a standard knowledge graph, specifically including:

[0043] Leveraging the API development capabilities of the work order management system, we integrate work order initiation channels, including the web, mobile app, customer service system, IoT device alarm interface, and email server interface, and record the access status of each channel.

[0044] Based on a unified channel access gateway, the data collection frequency is set to collect electronic work order source data from various channels in real time, forming a real-time work order data source set.

[0045] Based on the core entities and relationship chains in the initial knowledge graph, entity feature vectors are extracted and transformed into feature matrices.

[0046] Based on the feature matrix, feature information is extracted from each work order data in the real-time work order data source set and transformed into a feature vector to be classified.

[0047] Obtain the cosine similarity between the feature vector to be classified and the feature vectors of each entity category in the initial knowledge graph;

[0048] Based on cosine similarity, the work order data to be classified is divided into entity categories with the highest similarity and cosine similarity greater than or equal to the preset similarity threshold, thus obtaining the classified work order data set.

[0049] If the highest cosine similarity of the work order data is less than the preset similarity threshold, it is marked as needing manual classification. After manual classification by the operation and maintenance personnel, it is added to the classified work order data set.

[0050] Based on the classified work order data set, the demand features of each work order are extracted, and the feature data of processing resources are obtained from the initial knowledge graph.

[0051] Define a resource matching scoring function, and based on the resource matching scoring function, obtain the matching value of all candidate processing resources for each classified work order data, select the resource with the highest matching value that is greater than or equal to the preset scoring threshold as the optimal matching resource, and obtain the matching result;

[0052] Based on the matching results, the work order data is associated with the optimal matching resource information, and the fields corresponding to the work order are supplemented to obtain the standard source data of the electronic work order.

[0053] Based on standard source data from electronic work orders, the number of entities and relationships not covered in the initial knowledge graph were counted.

[0054] If the number of uncovered entities or the number of uncovered relationships is greater than or equal to the preset optimization trigger threshold, the new entities and relationships will be added to the initial knowledge graph, and the attribute values ​​of existing entities will be updated at the same time.

[0055] After the supplementation and update are completed, the optimized graph entity coverage rate is obtained to continue supplementing the uncovered entities and relationships until the optimized graph entity coverage rate is greater than or equal to the preset coverage rate threshold, thus obtaining the standard knowledge graph.

[0056] In an optional embodiment, the step of dynamically generating a list of preferred solutions and a knowledge push list based on a standard knowledge graph and combined with real-time collected electronic work order source data, followed by collecting work order processing flow information, specifically includes:

[0057] Based on the standard knowledge graph, extract the knowledge asset data related to work order processing, including historical solution sets, related SOP document sets, and historical case sets;

[0058] Based on real-time collected electronic work order source data, the key features of the current work order are extracted, including work order type, priority, fault description, and associated entities;

[0059] Obtain the feature similarity between the key features of the current work order and each solution in the historical solution set, combine the historical performance indicators of the solutions, define the solution priority function, and thus determine the solution priority value;

[0060] The historical solutions are sorted from highest to lowest priority value, and the solutions are filtered to form a list of preferred solutions.

[0061] From the set of associated SOP documents in the standard knowledge graph, SOP documents that match the key features of the current work order are selected to obtain a recommended set of SOPs.

[0062] From the historical case set, cases with a similarity to the key features of the current work order that is greater than or equal to a preset similarity threshold are selected to obtain a case recommendation set;

[0063] The recommended SOPs and case studies are arranged in order of priority: SOP documents first, case studies second, to form a knowledge push list;

[0064] The list of preferred solutions and the list of knowledge pushes are simultaneously pushed to the processing interface of the work order management system for processing personnel to view and select.

[0065] The staff handles work orders based on the list of preferred solutions and the knowledge push list. The work order management system records key information in real time during the process, including the start time, processing steps, resources called during the process, and end time.

[0066] When the work order processing status changes to "processed," the work order management system automatically generates a processing result questionnaire, pushes it to the user who initiated the work order, and sets a deadline for questionnaire collection.

[0067] If user feedback is received within the questionnaire return period, the feedback information will be integrated with the processing record.

[0068] If no feedback is received after the questionnaire return period, the issue will be marked as resolved by default, and the user's unanswered comment will be recorded.

[0069] Based on the integrated processing records and user feedback information, a complete work order processing workflow is formed.

[0070] In an optional embodiment, the step of archiving work orders based on work order processing flow information and performing a second optimization of the standard knowledge graph to obtain a preferred standard knowledge graph specifically includes:

[0071] Based on the work order processing flow information, extract the core identification information of the work order, including work order ID, processing status, processing resource ID, and user feedback result, as the core basis for archiving and classification;

[0072] Based on core identification information, work orders are archived and classified. Each category directory stores the corresponding work order processing flow information and associated electronic work order standard source data.

[0073] Get the number of work orders for each archive category. If the number of work orders for an archive category is greater than or equal to the preset compression threshold, then compress the files in that category.

[0074] If the number of work orders in the archived category is less than the preset compression threshold, the compression algorithm will be changed and the work orders will be compressed again until the number of work orders in the archived category is greater than or equal to the preset compression threshold, and the work order archiving will be completed.

[0075] Based on the archived work order processing flow information, the actual application effect of each entity-relationship-attribute combination in the standard knowledge graph is statistically analyzed, including the frequency of entity calls, the actual matching accuracy of relations, and the validity coefficient of attributes.

[0076] If there is an entity whose call frequency is less than or equal to the preset low-frequency threshold, it is marked as a low-frequency entity. Check whether the entity is a newly added entity in the business. If it is a newly added entity, it is retained.

[0077] If it is a redundant entity, it will be removed from the standard knowledge graph;

[0078] If there are relationships with actual matching accuracy rates lower than the preset accuracy threshold, analyze the reasons for the relationship matching errors and correct the relationship definition or matching rules.

[0079] If an attribute has a validity coefficient less than the preset validity threshold, it is determined to be a low-value attribute. If it is not a core attribute, it is removed from the attribute list of the corresponding entity.

[0080] Based on the results of low-frequency entity processing, relation correction, and attribute optimization, the entity set, relation set, and attribute set of the standard knowledge graph are updated. At the same time, new entities and new relations appearing in the archived work orders are added to the standard knowledge graph to obtain the optimized graph.

[0081] Obtain the accuracy of the optimized graph. If the accuracy of the optimized graph is less than the preset accuracy threshold, re-analyze the archived work order data, find erroneous entity-relationship-attribute combinations, and correct them until the accuracy of the optimized graph is greater than or equal to the preset accuracy threshold.

[0082] Simultaneously, the knowledge coverage rate of the optimized graph is obtained. If the knowledge coverage rate of the optimized graph is less than the preset knowledge coverage rate threshold, entities, relationships, and attributes of the business scenarios that are not covered are added until the knowledge coverage rate of the optimized graph is greater than or equal to the preset knowledge coverage rate threshold, thus obtaining the preferred standard knowledge graph.

[0083] Furthermore, a knowledge graph-based electronic management system for the entire work order process is proposed to implement any of the management methods mentioned above, including:

[0084] The work order domain ontology construction module is used to define the core entities of the work order management domain, thereby defining the relationships between the core entities, obtaining the core entity relationship chain, determining the attributes corresponding to each core entity, and constructing the work order management domain ontology graph.

[0085] The work order data acquisition and preprocessing module is used to acquire work order data sources based on the work order management domain ontology graph, and simultaneously divide the work order data sources into structured and unstructured data to obtain work order structured data and work order unstructured data. Then, the work order structured data and work order unstructured data are preprocessed respectively to obtain entity-relationship-attribute chain.

[0086] The initial knowledge graph construction and system integration module is used to construct an initial knowledge graph based on the entity-relationship-attribute chain, and then service the initial knowledge graph and integrate it with the work order management system.

[0087] The work order source data processing and knowledge graph first optimization module is used to integrate work order initiation channels using the work order management system, synchronously collect electronic work order source data in real time, classify the electronic work order source data according to the entity-relationship in the initial knowledge graph, and perform optimal resource matching on the classified electronic work order source data to obtain standard electronic work order source data, which is used to perform the first optimization of the initial knowledge graph to obtain a standard knowledge graph.

[0088] The work order knowledge application and knowledge graph secondary optimization module is used to dynamically generate a list of preferred solutions and a knowledge push list based on a standard knowledge graph and real-time collected electronic work order source data. It then collects work order processing flow information. Furthermore, it archives work orders based on this information, performs a second optimization of the standard knowledge graph to obtain a preferred standard knowledge graph, and finally interfaces the work order management system with this preferred standard knowledge graph.

[0089] In an optional embodiment, the work order domain ontology construction module includes:

[0090] The core entity and relationship definition unit is used to define the core entities in the work order management domain, thereby defining the relationships between the core entities and obtaining the core entity relationship chain.

[0091] A core entity attribute determination unit is used to determine the attributes corresponding to each core entity after obtaining the core entity relationship chain.

[0092] The work order domain ontology graph construction unit is used to construct a work order management domain ontology graph based on the core entity relationship chain and the attributes corresponding to each core entity.

[0093] In an optional embodiment, the work order data acquisition and preprocessing module includes:

[0094] A work order data source acquisition unit is used to acquire work order data sources based on the work order management domain ontology graph.

[0095] The work order data classification unit is used to synchronously divide the collected work order data sources into structured and unstructured data to obtain structured work order data and unstructured work order data.

[0096] The work order data preprocessing unit is used to preprocess the work order structured data and the work order unstructured data respectively to obtain the entity-relationship-attribute chain.

[0097] Compared with the prior art, the beneficial effects of the present invention are:

[0098] This solution proposes a knowledge graph-based electronic management method for the entire work order process. By defining core entities in the work order management domain, calculating the relationship weights based on the frequency of associations between entities, filtering core attributes, and defining semantic constraints, an ontology graph for the work order management domain is constructed. This achieves standardization and precision in the construction of the work order domain ontology, effectively avoiding the poor adaptability problem caused by the reliance on human experience in traditional ontology construction. It provides a unified semantic framework that fits the business scenario for work order data collection, processing, and knowledge graph construction, ensuring a standardized foundation for the entire work order process management.

[0099] This solution proposes a knowledge graph-based electronic management method for the entire work order process. It collects multi-source work order data based on the ontology graph of the work order management domain, divides the data into structured and unstructured data, and performs cleaning, standardization or textualization and entity matching processing on each data. The data is then fused to form an entity-relationship-attribute chain, achieving both completeness and efficiency in work order data processing. It takes into account the standardized processing of structured data to improve efficiency, while fully extracting implicit information from unstructured data, avoiding information fragmentation or loss, and providing high-quality, multi-dimensional data support for the initial knowledge graph construction.

[0100] This solution proposes a knowledge graph-based electronic management method for the entire work order process. It first optimizes the initial knowledge graph using standard source data from electronic work orders to obtain a standard knowledge graph. Then, by combining work order processing workflow information with the archived standard knowledge graph, a second optimization is achieved to obtain a preferred standard knowledge graph. Simultaneously, a list of preferred solutions and a knowledge push list are dynamically generated. This realizes a dynamic iteration of the knowledge graph and an intelligent closed loop for the entire work order process. The knowledge graph continuously improves its semantic association capabilities as work orders flow, enhancing the accuracy of resource matching and the success rate of work order processing. It also achieves closed-loop accumulation and reuse of work order knowledge, reducing the manual and time costs of subsequent work order processing and promoting continuous optimization of the work order management process. Attached Figure Description

[0101] Figure 1 This is a flowchart of a knowledge graph-based electronic management method for the entire process of work orders proposed in this invention.

[0102] Figure 2 This is a flowchart illustrating the construction process of the work order management ontology diagram in this invention.

[0103] Figure 3This is a flowchart illustrating the construction process of the entity-relationship-attribute chain in this invention.

[0104] Figure 4 This is a system framework diagram of a knowledge graph-based electronic management system for the entire work order process proposed in this invention. Detailed Implementation

[0105] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0106] Reference Figure 1 - Figure 4 As shown, a knowledge graph-based method for full-process electronic management of work orders includes:

[0107] Define the core entities in the work order management domain, thereby defining the relationships between the core entities and obtaining the core entity relationship chain. Then, determine the attributes corresponding to each core entity and construct the work order management domain ontology graph.

[0108] Based on the ontology graph of the work order management domain, work order data sources are collected and simultaneously divided into structured and unstructured data sources to obtain structured work order data and unstructured work order data. Then, data preprocessing is performed on the structured work order data and unstructured work order data respectively to obtain entity-relationship-attribute chain.

[0109] Based on the entity-relationship-attribute chain, an initial knowledge graph is constructed, and then the initial knowledge graph is serviced and integrated with the work order management system.

[0110] The work order management system is used to integrate work order initiation channels and collect electronic work order source data synchronously in real time. Then, the electronic work order source data is classified according to the entity-relationship in the initial knowledge graph, and the classified electronic work order source data is optimally matched with resources to obtain standard electronic work order source data. This data is used to perform the first optimization of the initial knowledge graph to obtain the standard knowledge graph.

[0111] Based on the standard knowledge graph and combined with real-time collected electronic work order source data, a list of preferred solutions and a knowledge push list are dynamically generated, and then work order processing flow information is collected.

[0112] Based on the work order processing flow information, the work orders are archived, and the standard knowledge graph is optimized a second time to obtain the preferred standard knowledge graph. Then, the work order management system is connected with the preferred standard knowledge graph.

[0113] Furthermore, the core entities of the work order management domain are defined, thereby defining the relationships between the core entities and obtaining the core entity relationship chain. Then, the attributes corresponding to each core entity are determined, and the ontology graph of the work order management domain is constructed, specifically including:

[0114] Based on the business scenario requirements of the work order management field, we conducted domain research to obtain business process documents, historical work order records, and domain expert experience data related to work order management.

[0115] Based on business process documents, historical work order records, and domain expert experience data, a candidate set of entities that appear frequently and play a key supporting role in the work order flow is selected. The criterion for high frequency is that the frequency of the entity in the historical work order records is F≥F0 (F0 is a preset threshold, F0=total number of historical work orders×10%, and if the calculation result is a decimal, it is rounded up).

[0116] Based on the candidate entity set, domain experts are invited to score the coreity of the candidate entities. The coreity score S is based on a 1-5 scale (5 points is the highest coreity). Entities with a coreity score S≥3 are selected to obtain the core entity list of the work order management domain. Core entities include, but are not limited to, work orders, users, fault types, processing resources, process nodes, and knowledge assets.

[0117] Based on the core entity list, the interaction scenarios of each core entity in the business process are analyzed, the association frequency F_ij between entities is counted (i.e., the number of times entity i and entity j appear together in the business process), and the relationship weight W_ij between entities is calculated. The calculation formula is W_ij=(F_ij / F_total)×K_i (where F_total is the total association frequency between all entities, and K_i is the business importance coefficient of entity i. K_i is set by domain experts according to the degree of influence of the entity on the work order processing, and the value range is 1-3).

[0118] Based on the relation weight W_ij, entity relations with W_ij≥0.2 are selected (0.2 is the preset relation weight threshold), and sorted from high to low according to the frequency of association to obtain the core entity relation chain;

[0119] Based on the core entity relationship chain and the business scenario requirements of the core entities, determine the attribute items corresponding to each core entity. The attribute items must cover the key features of the entity. For example, the attributes of the work order entity must include work order ID, creation time, priority, and status. At the same time, calculate the completeness requirement coefficient C for each attribute item (C = number of records with no missing items in historical work orders / total number of historical work order records), and select attribute items with C≥0.9 as core attributes.

[0120] Based on the core entity list, core entity relationship chain, and core entity core attributes, ontology modeling is performed using ontology building tools (such as Protégé), defining semantic constraints on entities, relationships, and attributes, and generating an ontology graph for the work order management domain.

[0121] Based on the generated work order management domain ontology graph, consistency verification is performed by domain experts. The verification pass rate P = (number of entity-relationship-attribute combinations that experts agree on / total number of entity-relationship-attribute combinations in the ontology graph) × 100%. When P ≥ 95%, the ontology graph is determined to be the final work order management domain ontology graph.

[0122] If P < 95%, then the entities, relationships, or attributes are corrected based on expert feedback, and consistency checks are performed again until the pass rate is ≥ 95%.

[0123] Understandably, by conducting domain research to obtain multi-source data, selecting core entities based on high frequency and coreity scores, calculating relationship weights based on association frequency and business importance to select core relationships, determining core attributes based on integrity requirement coefficients, and then constructing a work order management domain ontology graph through ontology modeling and expert consistency verification (pass rate ≥95%), the standardization, accuracy, and business fit of work order domain ontology construction are achieved. This approach ensures the scientific nature of core entities, relationships, and attributes through a combination of data-driven and expert experience (such as selecting entities based on both high frequency and coreity), while guaranteeing the semantic accuracy and logical rigor of the ontology graph through weight calculation and consistency verification. This provides a unified and business-scenario-appropriate semantic framework for subsequent work order data processing, knowledge graph construction, and end-to-end management, effectively avoiding the problems of poor adaptability and logical confusion caused by relying on manual experience in traditional ontology construction.

[0124] Furthermore, based on the ontology graph of the work order management domain, work order data sources are collected and simultaneously divided into structured and unstructured data sources to obtain structured work order data and unstructured work order data. Then, data preprocessing is performed on the structured and unstructured work order data respectively to obtain the entity-relationship-attribute chain, specifically including:

[0125] Based on the ontology diagram of the work order management domain, the collection scope of work order data sources is determined. The collection scope includes scanned copies of historical paper work orders, historical electronic work orders in the work order management system, business data from business-related systems (such as CRM systems and operation and maintenance monitoring systems), and processing manuals and training documents written by domain experts, forming the original work order data source set D.

[0126] Based on the original work order data source set D, the data format characteristics of each data source are extracted. The data format characteristics include whether the fields are fixed, whether the data structure is standardized, and whether it contains free text. The matching degree M between the data source and the core fields of the ontology diagram is calculated. The calculation formula is M = (number of fields in the data source that match the core fields of the ontology diagram / total number of core fields in the ontology diagram) × 100%.

[0127] Based on the matching degree M and data format characteristics, the original work order data source set D is divided into work order structured data D1 and work order unstructured data D2. When M≥80% and the data format is fixed and standardized, it is determined to be work order structured data D1.

[0128] When M < 80% or contains free text or has an inconsistent data format, it is determined to be unstructured work order data D2;

[0129] Based on the work order structure data D1, a data cleaning algorithm is used to remove redundant data (the criteria for redundancy data is records with ≥3 duplicate records) and abnormal data (the criteria for abnormal data is numerical attributes that exceed the range of [μ-3σ,μ+3σ], where μ is the mean of the attribute and σ is the standard deviation of the attribute), resulting in cleaned structured data D11.

[0130] Based on the cleaned structured data D11, data standardization is performed according to the attribute definition of the ontology graph, unifying the same attribute value in different formats into a preset format (such as unifying the date attribute into the format "YYYY-MM-DDHH:MM:SS"), resulting in standardized structured data D12;

[0131] Based on the unstructured work order data D2, OCR technology is used to recognize the text content in the scanned paper work order, and speech-to-text technology is used to process the speech data source to obtain textual unstructured data D21.

[0132] Based on the textual unstructured data D21, word segmentation tools (such as jieba) are used for word segmentation to remove stop words (the stop word list adopts the general Chinese stop word list), and the entity information in the text is extracted using the BERT+CRF entity recognition model and matched with the core entities in the ontology graph to obtain the entity matching result R.

[0133] Based on the entity matching result R, the relationship between entities is analyzed in conjunction with the text context. Referring to the core entity relationship chain in the ontology graph, the entity-relationship pairs in the unstructured data are determined. At the same time, the attribute information corresponding to the entity is extracted (e.g., the "creation time" attribute of the work order entity is extracted as "2025-10-10" and the "priority" attribute is "P0" from "P0 level work order created on October 10, 2025"), and the entity-relationship-attribute set S2 of the unstructured data is obtained.

[0134] Based on the standardized structured data D12, the core entities, relations and attributes in the ontology graph are directly mapped to obtain the entity-relation-attribute set S1 of the structured data;

[0135] Based on the entity-relationship-attribute set S1 of structured data and the entity-relationship-attribute set S2 of unstructured data, data fusion is performed to eliminate duplicate entity-relationship-attribute combinations (the duplicate judgment criteria are that the entity ID, relation type, and attribute value are completely identical) to obtain the final entity-relationship-attribute chain L.

[0136] Understandably, by determining the data collection scope of multi-source work order data (including historical paper / electronic work orders, business-related system data, and expert manuals) based on the ontology graph of the work order management domain, an original set D is formed. This is then divided into structured data D1 (M≥80% and formatted correctly) and unstructured data D2 based on the matching degree M (M=number of matching fields / total number of core fields × 100%) and data format characteristics. D1 is then cleaned (removing redundant data ≥3 times and abnormal data [μ-3σ]) and standardized. D2 is processed through OCR / speech-to-text conversion, word segmentation and stop word removal, and BERT+CRF entity recognition to match the ontology graph, extracting entity-relationship-attribute sets. Finally, the structured and unstructured entity-relationship-attribute sets are merged and duplicates are removed to obtain the entity-relationship-attribute chain L. This achieves full-dimensional collection, accurate classification processing, and high-quality integration of work order data. This approach not only covers all scenarios of work order management data through multi-source collection, avoiding the problem of missing key information in traditional collection methods, but also treats structured and unstructured data differently, ensuring the standardization and efficiency of structured data while fully mining implicit information in unstructured data (such as work order attributes in free text). The resulting entity-relationship-attribute chain L data is of high quality and has clear semantic associations, providing full-dimensional and reliable data support for the accurate construction of the initial knowledge graph. It solves the problems of insufficient consideration of both structured and unstructured data and data disorder that cannot be directly reused in traditional data processing.

[0137] Furthermore, based on the entity-relationship-attribute chain, an initial knowledge graph is constructed. This initial knowledge graph is then service-oriented and integrated with the work order management system, specifically including:

[0138] Based on the entity-relationship-attribute chain L, analyze the number of entities N, the number of relations M, and the number of attributes P in the chain, and calculate the graph storage requirement S. The calculation formula is S = (N × average entity storage size + M × average relation storage size + P × average attribute storage size) × 1.2 (1.2 is the reserved storage redundancy coefficient).

[0139] Based on the storage requirement size S, select a suitable graph database (such as Neo4j or NebulaGraph). If S≤100GB, select the lightweight graph database Neo4j.

[0140] If S > 100GB, choose the distributed graph database NebulaGraph;

[0141] Based on the selected graph database, create a knowledge graph template schema, define entity tags (corresponding to core entities), relation types (corresponding to core entity relations), and attribute keys (corresponding to core entity attributes), and ensure that the schema is consistent with the structure of the entity-relationship-attribute chain L;

[0142] Based on the created knowledge graph template schema, the data in the entity-relationship-attribute chain L is imported into the graph database in batches. The amount of data B imported in each batch is set to 80% of the maximum amount of data that can be imported in a single batch in the graph database (the maximum amount of data imported is obtained through database performance testing). During the import process, the data import success rate R_import is monitored in real time. The calculation formula is R_import = (number of successfully imported entity-relationship-attribute combinations / total number of imported entity-relationship-attribute combinations) × 100%. If R_import < 99%, the import is paused, and the batch of data is re-imported after checking for data format problems.

[0143] After all the data is imported, entity indexes and relationship indexes are built. The index fields include entity ID and core attributes (such as work order ID and user ID) to improve graph query efficiency. The query response time T_query after the index is built is tested, and T_query ≤ 500ms is required (the query test uses 100 high-frequency query statements and takes the average response time).

[0144] If T_query > 500ms, optimize the index structure (e.g., add a composite index), retest until T_query ≤ 500ms, and obtain the initial knowledge graph G0;

[0145] Based on the initial knowledge graph G0, the graph service interface is developed using the Spring Boot framework. It encapsulates three core APIs: query (e.g., querying relationships by entity ID), add (e.g., adding entity-relationship-attribute combinations), and update (e.g., updating entity attribute values). It defines the request parameter format, response parameter format, and error codes (e.g., error code 400 indicates an error in the request parameters, and 500 indicates an internal server error).

[0146] Based on the developed graph service API, write API test cases. The test cases cover normal requests and abnormal requests (such as missing parameters or incorrect parameter formats). The total number of test cases is C = number of API types × 5 (5 test cases for each API). Execute the test cases and calculate the API test pass rate R_api = (number of passed test cases / total number of test cases) × 100%. R_api ≥ 98% is required.

[0147] If R_api < 98%, fix the vulnerabilities in the API code and retest until R_api ≥ 98%, then complete the service-oriented transformation of the initial knowledge graph.

[0148] Obtain the API documentation of the work order management system and analyze the system's API types (such as RESTAPI, SOAPAPI), request protocols (such as HTTP, HTTPS), and data interaction formats (such as JSON, XML).

[0149] Based on the interface information of the work order management system, an adapter for the graph service and the work order management system was developed to realize data format conversion and protocol adaptation between the two.

[0150] After the integration is completed, joint debugging tests are conducted to simulate the scenario of the work order management system calling the graph service API (such as querying the fault type associated with the work order). A total of 100 joint debugging tests are conducted. The success rate of joint debugging is calculated as R_link = (number of successful joint debugging tests / total number of joint debugging tests) × 100%, and R_link ≥ 95% is required.

[0151] If R_link < 95%, check the compatibility of the interface adapter, re-integrate until R_link ≥ 95%, and complete the initial integration of the knowledge graph with the work order management system.

[0152] Understandably, the graph storage requirement S is calculated based on the entity-relationship-attribute chain L (S = (number of entities N × average entity storage size + number of relations M × average relation storage size + number of attributes P × average attribute storage size) × 1.2, where 1.2 is the redundancy coefficient). A graph database is selected based on S (Neo4j for S≤100GB, NebulaGraph for S>100GB). After creating a schema consistent with the chain L structure, data is imported in batches (80% of the maximum import volume per batch), and the import success rate R_import ≥ 1. With a success rate of 99%, the initial knowledge graph G0 was obtained by building an index and testing the query response time T_query≤500ms. Next, three types of APIs (query, add, update) were developed using Spring Boot. Test cases were designed according to C=number of API types×5 to ensure that the API test pass rate R_api≥98%, thus completing the service-oriented architecture. Finally, the interface of the work order system was analyzed and a docking adapter was developed. After 100 joint debugging sessions, the success rate of the joint debugging was ensured to be R_link≥95%, thus completing the docking. This achieved the adaptation and construction of the initial knowledge graph, efficient service-oriented architecture, and stable integration with the work order management system. By adapting storage selection to meet specific needs, it avoids the resource waste or performance deficiencies caused by blindly choosing databases in traditional methods. Furthermore, through data import monitoring and index optimization, it ensures the integrity of graph data (R_import≥99%) and query efficiency (T_query≤500ms). Additionally, through API testing and integration verification, it ensures the reliability of the graph service (R_api≥98%) and seamless system integration (R_link≥95%). This effectively solves the problems of storage mismatch, data loss, slow queries, and unstable integration with business systems after service-oriented architecture in traditional graph construction, providing stable knowledge graph service support for subsequent intelligent processes such as work order classification and resource matching.

[0153] Furthermore, the work order management system is used to integrate work order initiation channels and synchronously collect electronic work order source data in real time. Then, the electronic work order source data is classified according to the entity-relationship structure in the initial knowledge graph. Optimal resource matching is then performed on the classified electronic work order source data to obtain standard electronic work order source data. This standard knowledge graph is used for the first optimization of the initial knowledge graph, resulting in a standard knowledge graph, which specifically includes:

[0154] Leveraging the API development capabilities of the work order management system, integrate work order initiation channels, including the web, mobile app, customer service system, IoT device alarm interface, and email server interface, to form a unified channel access gateway. Record the access status (online / offline) of each channel and monitor the channel access success rate in real time: R_channel = (number of successfully connected channels / total number of integrated channels) × 100%, requiring R_channel ≥ 99%.

[0155] If R_channel < 99%, check the channel interface configuration and reconnect to the offline channel until R_channel ≥ 99%;

[0156] Based on a unified channel access gateway, the data collection frequency f is set. The collection frequency f is set according to the number of work orders generated by each channel. For channels with a work order generation volume of ≥10 / minute, f=1 second / time.

[0157] For channels with a work order generation rate of <10 orders / minute, f=5 seconds / time, collect electronic work order source data from each channel in real time to form a real-time work order data source set D_source;

[0158] Based on the core entity-relationship chain in the initial knowledge graph G0, entity feature vectors are extracted (such as the feature vector of a work order entity, which includes work order type, priority, and associated device ID). The TF-IDF algorithm is then used to transform the entity feature vectors into feature matrices.

[0159] Based on the feature matrix, feature information is extracted from each work order data in the real-time work order data source set D_source and transformed into a feature vector V to be classified;

[0160] Calculate the cosine similarity Sim between the feature vector V to be classified and the feature vectors V0 of each entity category in the initial knowledge graph G0. The calculation formula is Sim=(V・V0) / (|V|×|V0|), where "・" represents the vector dot product, and |V| and |V0| represent the magnitude of the vectors.

[0161] Based on cosine similarity Sim, the work order data to be classified is divided into the entity category with the highest similarity and Sim≥0.7 (0.7 is the preset similarity threshold), resulting in the classified work order data set D_class; if the highest Sim of a certain work order data is <0.7, it is marked as "to be manually classified", and added to D_class by the operation and maintenance personnel after manual classification.

[0162] Based on the classified work order data set D_class, the requirement features of each work order (such as fault type and processing time limit requirements) are extracted. At the same time, feature data of processing resources (personnel / team) (such as skill tags, current load, and historical processing success rate) are obtained from the initial knowledge graph G0.

[0163] Define the resource matching scoring function as Score = α × S_skill + β × S_load + γ × S_success, where α, β, and γ are weighting coefficients (α + β + γ = 1, and α = 0.5, β = 0.3, γ = 0.2, set by domain experts), S_skill is the skill matching degree (number of skill matches / total number of skills required for the work order), S_load is the load coefficient (1 - current load / maximum load, where maximum load is the maximum number of work orders that the resource can process per unit time), and S_success is the historical processing success rate (number of times this type of work order was successfully processed in the past / total number of times this type of work order was processed in the past).

[0164] Based on the resource matching scoring function Score, for each categorized work order data, the Score value of all candidate processing resources is calculated, and the resource with the highest Score value and a Score ≥ 80 (80 is the preset scoring threshold) is selected as the optimal matching resource, and the matching result R_match is obtained.

[0165] Based on the matching result R_match, the work order data is associated with the optimal matching resource information, and the fields such as "processing resource ID" and "expected processing time" (expected processing time = historical average processing time of this type of work order × S_load) of the work order are supplemented to obtain the electronic work order standard source data D_standard;

[0166] Based on the electronic work order standard source data D_standard, the number of entities N_new (i.e., the number of entities that appear in D_standard but do not exist in G0) and the number of relationships M_new (i.e., the number of relationships that appear in D_standard but do not exist in G0) that are not covered in the initial knowledge graph G0 are counted.

[0167] If N_new≥5 or M_new≥3 (5 and 3 are preset optimization trigger thresholds), then the new entity and new relationship will be added to the initial knowledge graph G0, and the attribute values ​​of existing entities will be updated (such as updating the "historical processing success rate" of the processing resource).

[0168] After the supplementation and update are completed, calculate the optimized map entity coverage C_cover = (number of entities in the map / total number of entities in D_standard) × 100%, requiring C_cover ≥ 85%;

[0169] If C_cover < 85%, continue to supplement the uncovered entities and relations until C_cover ≥ 85%, and obtain the standard knowledge graph G1.

[0170] Understandably, by utilizing the work order management system to integrate initiation channels such as the web terminal, mobile APP, customer service system, IoT device alarm interface, and email server interface, a unified channel access gateway is formed, and the channel access success rate R_channel = (number of successfully accessed channels / total number of integrated channels) × 100% (requiring R_channel ≥ 99%, if not reaching the target, a reconnection check is performed); based on the gateway, the collection frequency f is set according to the channel work order generation volume (≥10 orders / minute = 1 second / time, <10 orders / minute = 5 seconds / time), and the electronic work order source data set D_source is collected in real time; then, based on the core entity-relationship chain of the initial knowledge graph G0, feature vectors are extracted, and TF-IDF is used to convert them into feature matrices. The cosine similarity Sim between the work order to be classified vector V and the entity category vector V0 is calculated (Sim = (V・V0) / (|V|×|V0)). The data is categorized according to Sim≥0.7 (with manual input for <0.7) to obtain D_class. Then, the work order requirement features and resource features (skills, load, historical success rate) are extracted. The optimal resource with a Score≥80 is selected using the scoring function Score=0.5S_skill+0.3S_load+0.2S_success (α+β+γ=1), and the supplementary fields are associated to obtain the electronic work order standard source data D_standard. Finally, the number of entities N_new and relations M_new not covered by G0 in D_standard are counted. When N_new≥5 or M_new≥3, updates are made to ensure the entity coverage rate C_cover≥85%, resulting in the standard knowledge graph G1. This achieves unified work order channels, real-time data collection, accurate work order classification, optimized resource matching, and dynamic optimization of the initial knowledge graph. It addresses the issue of missing work orders caused by the fragmentation of traditional channels through multi-channel integration and a high access success rate (≥99%), and ensures data real-time performance by dynamically setting the collection frequency based on the number of work orders. It also avoids the problems of fuzzy and misclassified work orders in traditional systems by combining cosine similarity classification with manual supplementation. Furthermore, it solves the problems of blind resource allocation and uneven load by selecting the optimal resources through multi-dimensional scoring functions. Finally, it optimizes the graph based on standard source data to solve the defects of the initial graph being static and lacking business coverage, providing knowledge support that is more in line with business needs for subsequent intelligent processing of work orders.

[0171] Furthermore, based on the standard knowledge graph and combined with real-time collected electronic work order source data, a list of preferred solutions and a knowledge push list are dynamically generated. Then, work order processing flow information is collected, specifically including:

[0172] Based on the standard knowledge graph G1, extract the knowledge asset data related to work order processing, including the historical solution set S_sol (including the number of times each solution was applied C_sol, the number of successful processing C_success, and the average processing time T_avg), the associated SOP document set S_sop, and the historical case set S_case.

[0173] Based on the real-time collected electronic work order source data D_source, the key features F_current of the current work order are extracted, including work order type T, priority P, fault description D_fault, and associated entities E (such as associated equipment and associated users).

[0174] Calculate the feature similarity Sim_sol between the current work order's key feature F_current and each solution in the historical solution set S_sol. Sim_sol is calculated by first converting F_current and the solution features into vectors, and then calculating the cosine similarity, using the same formula as Sim in dependent claim 5.

[0175] Based on the similarity Sim_sol and combined with the historical performance indicators of the scheme, the scheme priority function is defined as Priority=Sim_sol×(C_success / C_sol)×(1 / T_avg), where (C_success / C_sol) is the scheme success rate and (1 / T_avg) is the processing efficiency coefficient.

[0176] The historical solutions are sorted from highest to lowest Priority value, and the top K solutions with the highest Priority value are selected (K is set according to the work order priority: K=5 for P0 level, K=4 for P1 level, K=3 for P2 level, and K=2 for P3 level), forming a list of preferred solutions L_sol;

[0177] Based on the current work order's key feature F_current, SOP documents that match F_current are selected from the associated SOP document set S_sop in the standard knowledge graph G1 (the matching criteria are that the applicable work order type of the SOP document contains T and the applicable fault scenario contains the keywords in D_fault), resulting in the SOP recommendation set S_sop_rec.

[0178] From the historical case set S_case, select cases with a similarity Sim_case ≥ 0.8 to the key feature F_current of the current work order (Sim_case is calculated in the same way as Sim_sol) to obtain the case recommendation set S_case_rec; arrange the SOP recommendation set S_sop_rec and the case recommendation set S_case_rec in the order of "SOP document first, case second" to form the knowledge push list L_know;

[0179] The preferred solution list L_sol and the knowledge push list L_know are simultaneously pushed to the processing interface of the work order management system for processing personnel to view and select. Processing personnel carry out work order processing based on L_sol and L_know. The work order management system records key information in real time during the processing process, including the processing start time T_start, the processing step record S_step (including the execution time T_step and execution result R_step of each step), the resources called during the processing R_use (such as the tools called and the collaborating personnel), and the processing end time T_end.

[0180] When the work order processing status changes to "processed", the work order management system automatically generates a processing result questionnaire and pushes it to the user who initiated the work order. The questionnaire includes three core questions: "Is the problem resolved?" (Yes / No), "Processing satisfaction rating" (1-5 points), and "Additional feedback" (text box). The questionnaire return time limit T_limit is set (P0 level T_limit=1 hour, P1 level T_limit=2 hours, P2 level T_limit=4 hours, P3 level T_limit=8 hours).

[0181] If user feedback is received within T_limit, the feedback information and processing record will be integrated; if no feedback is received after T_limit, the problem will be marked as "problem solved" by default, the satisfaction score will be calculated as 3 points, and the note "no user feedback" will be recorded.

[0182] Based on the integrated processing records (T_start, S_step, R_use, T_end) and user feedback information, a complete work order processing flow information I_process is formed, where the processing time T_process = T_end - T_start, and the processing success rate R_process = 1 (if the user feedback "the problem has been solved" or the default mark "the problem has been solved") or 0 (if the user feedback "the problem has not been solved").

[0183] Understandably, by extracting the historical solution set S_sol (including application count C_sol, success count C_success, and average duration T_avg) from the standard knowledge graph G1, the associated SOP document set S_sop, and the historical case set S_case, and extracting the current work order's key features F_current (type T, priority P, fault description D_fault, and associated entity E) based on the real-time electronic work order source data D_source, the feature similarity Sim_sol (cosine similarity) between F_current and the solutions in S_sol is calculated. Combining the solution success rate (C_success / C_sol) and the processing efficiency coefficient (1 / T_avg), a priority function Priority = Sim_sol × (C_success / C_sol) × (1 / T_avg) is defined. The top K solutions are then selected based on the Priority value (P0 level K=5, P1 level K=4, P2 level K=3). (P3 level K=2) to form a preferred solution list L_sol; then filter SOP documents (applicable types containing T, fault scenarios containing D_fault keywords) that match F_current and cases with similarity Sim_case≥0.8, and form a knowledge push list L_know according to "SOP first, case second", and push both lists to the work order system; at the same time, record key processing information (T_start, S_step, R_use, T_end), and push a questionnaire containing 3 core questions after the work order is completed (P0 level return time limit of 1 hour, increasing according to priority). If no feedback is received within the time limit, it is assumed that "the problem has been solved" and 3 points are scored. Finally, integrate the records and feedback to form work order processing process information I_process containing processing time T_process=T_end-T_start and processing success rate R_process, realizing accurate matching and push of work order processing knowledge, full-dimensional traceability of the processing process, and standardized closed-loop management of user feedback. It addresses the issues of unfounded traditional solution recommendations and a disconnect between knowledge delivery and work order requirements by employing multi-dimensional priority functions and similarity filtering, enabling personnel to quickly access highly adaptable solutions and knowledge, thus improving processing efficiency. Furthermore, it avoids the shortcomings of a lack of traceability and difficulty in subsequent tracking through full-process information recording. Finally, it solves the problems of untimely user feedback collection and inconsistent standards by setting questionnaire return deadlines and feedback fallback rules based on priority, providing complete and reliable process data support for subsequent work order archiving and graph optimization.

[0184] Furthermore, based on the work order processing flow information, work orders are archived, and the standard knowledge graph is optimized a second time to obtain a preferred standard knowledge graph, specifically including:

[0185] Based on the work order processing flow information I_process, extract the core identification information of the work order, including work order ID, processing status (resolved / unresolved), processing resource ID, and user feedback result, as the core basis for archiving and classification;

[0186] Based on core identification information, work orders are archived and classified according to the hierarchical structure of "processing status → work order type → processing time (year and month)", such as "resolved → IT operation and maintenance → 202510". Each category directory stores the corresponding work order processing flow information I_process and the associated electronic work order standard source data D_standard.

[0187] Calculate the number of work orders C_archive for each archive category. If C_archive ≥ 100 (100 is the preset compression threshold), then use the ZIP compression algorithm to compress the files in that directory. The compression ratio R_compress = (total file size before compression - total file size after compression) / total file size before compression × 100%, and R_compress ≥ 50% is required.

[0188] If R_compress < 50%, then switch to the 7Z compression algorithm and recompress until R_compress ≥ 50%, then complete the work order archiving;

[0189] Based on the archived work order processing flow information I_process, the actual application effect of each entity-relationship-attribute combination in the standard knowledge graph G1 is statistically analyzed, including the frequency of entity calls F_call (the number of times a certain entity appears in the archived work orders), the actual matching accuracy of relations R_rel (the number of work orders that are successfully matched with resources based on this relation / the total number of work orders that are matched with resources based on this relation), and the validity coefficient of attributes C_attr (the success rate of work order processing with this attribute / the success rate of work order processing without this attribute).

[0190] If an entity's F_call ≤ 10 (10 is the preset low-frequency threshold), it is marked as a "low-frequency entity". Check if the entity is a newly added entity for the business. If it is a newly added entity and may be reused in the future, it is retained.

[0191] If it is a redundant entity, it will be removed from G1;

[0192] If R_rel < 60% (60% is the preset accuracy threshold) of a relation, analyze the reasons for the relation matching error, correct the relation definition or matching rules, such as adjusting the relation weight calculation method, and recalculate the matching accuracy of the relation until R_rel ≥ 60%;

[0193] If the C_attr of a certain attribute is less than 1.2 (1.2 is the preset validity threshold, that is, the processing success rate of the attribute must be more than 20% higher than that of the attribute without the attribute), then the attribute is determined to be a "low-value attribute". If it is not a core attribute, it is removed from the attribute list of the corresponding entity.

[0194] Based on the results of low-frequency entity processing, relation correction, and attribute optimization, the entity set, relation set, and attribute set of the standard knowledge graph G1 are updated. At the same time, new entities (such as new types of faults) and new relations (such as new collaborative relations) appearing in archived work orders are added to G1 to obtain the optimized graph G1_opt.

[0195] Calculate the graph accuracy R_acc of G1_opt = (number of correct entity-relation-attribute combinations / total number of entity-relation-attribute combinations in G1_opt) × 100%, and require R_acc ≥ 92%;

[0196] If R_acc < 92%, then re-analyze the archived work order data, find incorrect entity-relationship-attribute combinations, and correct them until R_acc ≥ 92%.

[0197] Simultaneously calculate the knowledge coverage rate of G1_opt, C_know = (number of business scenarios covered by G1_opt / total number of business scenarios) × 100%, requiring C_know ≥ 90%;

[0198] If C_know < 90%, then supplement the entities, relationships, and attributes that are not covered in the business scenarios until C_know ≥ 90%, and obtain the preferred standard knowledge graph G2.

[0199] Understandably, the core identification information of the work order (work order ID, processing status, processing resource ID, user feedback result) is extracted based on the work order processing flow information I_process. This information is then archived according to the hierarchical structure of "processing status → work order type → processing time (year / month)," storing the corresponding I_process and the electronic work order standard source data D_standard. The number of work orders in the archived directory C_archive is calculated. When C_archive ≥ 100, ZIP compression is used (requiring a compression rate R_compress ≥ 50%; if not, 7Z is used to complete the archiving). Finally, based on the archived I_process, the entity call frequency F_call and the relationship matching accuracy R_rel (number of successfully matched work orders / total number of matched work orders) in the standard knowledge graph G1 are statistically analyzed. The system calculates the number of singular entities and the attribute validity coefficient C_attr (success rate of processing with and without attributes). It processes low-frequency entities (F_call≤10, distinguishing between newly added and redundantly deleted entities), corrects low-accuracy relationships (R_rel<60%, adjusting definitions / rules to R_rel≥60%), removes low-value attributes (C_attr<1.2, removing non-core attributes), supplements archived work orders with new entities / relationships, and finally ensures that the optimized graph accuracy R_acc = (number of correct combinations / total number of combinations) × 100% ≥ 92% and the knowledge coverage C_know = (number of covered scenarios / total number of scenarios) × 100% ≥ 90%, resulting in the preferred standard knowledge graph G2. This achieves standardized and efficient work order archiving and dynamic, precise, iterative optimization of the knowledge graph. It addresses the problems of disordered and difficult-to-retrieve traditional work order archiving and storage redundancy through hierarchical archiving and conditional compression (quantity threshold 100, compression rate ≥50%), making archived data easy to trace and saving storage resources. Furthermore, it solves the defects of traditional knowledge graphs caused by static solidification, such as redundant entity relationships, low matching accuracy, low attribute value, and insufficient business coverage, through graph optimization based on actual application results (statistical analysis of F_call, R_rel, and C_attr and targeted processing). This ensures that the graph continuously meets the business needs of work order management, providing more accurate, comprehensive, and efficient knowledge support for subsequent intelligent processing of the entire work order process, forming a closed-loop value cycle of "work order processing - archiving - graph optimization".

[0200] Furthermore, a knowledge graph-based electronic management system for the entire work order process is proposed to implement any of the management methods mentioned above, including:

[0201] The work order domain ontology construction module is used to define the core entities of the work order management domain, thereby defining the relationships between the core entities, obtaining the core entity relationship chain, determining the attributes corresponding to each core entity, and constructing the work order management domain ontology graph.

[0202] The work order data acquisition and preprocessing module is used to acquire work order data sources based on the ontology graph of the work order management domain. Simultaneously, the work order data sources are divided into structured and unstructured data to obtain structured work order data and unstructured work order data. Then, the structured work order data and unstructured work order data are preprocessed respectively to obtain entity-relationship-attribute chain.

[0203] The initial knowledge graph construction and system integration module is used to construct an initial knowledge graph based on the entity-relationship-attribute chain, and then service-ize the initial knowledge graph and integrate it with the work order management system.

[0204] The first optimization module for work order source data processing and knowledge graph is used to integrate work order initiation channels using the work order management system, synchronously and in real time collect electronic work order source data, then classify the electronic work order source data according to the entity-relationship in the initial knowledge graph, and perform optimal resource matching on the classified electronic work order source data to obtain standard electronic work order source data, which is used to perform the first optimization of the initial knowledge graph to obtain a standard knowledge graph.

[0205] The work order knowledge application and knowledge graph secondary optimization module is used to dynamically generate a list of preferred solutions and a knowledge push list based on a standard knowledge graph and real-time collected electronic work order source data. It then collects work order processing flow information; furthermore, it archives work orders based on this information, performs a second optimization of the standard knowledge graph to obtain a preferred standard knowledge graph, and finally interfaces the work order management system with this preferred standard knowledge graph.

[0206] In an optional embodiment, the work order domain ontology construction module includes:

[0207] The core entity and relationship definition unit is used to define the core entities in the work order management domain, thereby defining the relationships between the core entities and obtaining the core entity relationship chain.

[0208] The core entity attribute determination unit is used to determine the attributes corresponding to each core entity after obtaining the core entity relationship chain.

[0209] The work order domain ontology graph construction unit is used to construct the work order management domain ontology graph based on the core entity relationship chain and the attributes corresponding to each core entity.

[0210] In an optional embodiment, the work order data acquisition and preprocessing module includes:

[0211] The work order data source acquisition unit is used to collect work order data sources based on the work order management domain ontology graph.

[0212] The work order data classification unit is used to synchronously divide the collected work order data sources into structured and unstructured data to obtain structured work order data and unstructured work order data.

[0213] The work order data preprocessing unit is used to preprocess the structured and unstructured work order data to obtain the entity-relationship-attribute chain.

[0214] In an optional embodiment, the initial knowledge graph construction and system integration module includes:

[0215] The initial knowledge graph construction unit is used to construct an initial knowledge graph based on the entity-relationship-attribute chain.

[0216] The initial knowledge graph service unit is used to service the initial knowledge graph after it has been constructed.

[0217] The knowledge graph and work order system docking unit is used to dock the initial knowledge graph after service-oriented processing with the work order management system.

[0218] The advantages of this invention are as follows: It constructs a standardized work order domain ontology, combining multi-source data with quantitative screening rules (such as entity frequency and coreity scores) and expert verification, avoiding reliance on manual experience, providing a unified semantic framework, and laying the foundation for standardized management; it processes data in all dimensions, differentially cleaning standardized structured data and extracting implicit information from unstructured data, balancing efficiency and information integrity, providing high-quality support for the knowledge graph; the knowledge graph dynamically iterates, undergoing two optimizations to adapt to business changes and resolve static defects; the work order process is intelligent, integrating multiple channels, automatically classifying and matching optimal resources, pushing solutions and knowledge, and improving processing efficiency; it forms a knowledge closed loop, with archived information feeding back into the knowledge graph, realizing knowledge reuse and promoting continuous optimization of work order management.

[0219] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A knowledge graph-based work order full-process electronic management method, characterized in that, The application relates to a method for constructing a knowledge graph of a work order management field. The method comprises the following steps: defining core entities of the work order management field, thereby defining relationships between the core entities, obtaining a core entity relationship chain, and then determining attributes corresponding to each core entity to construct a work order management field ontology graph; based on the work order management field ontology graph, collecting work order data sources, synchronously dividing the work order data sources into structured and unstructured data, obtaining work order structure data and work order unstructured data, and then respectively performing data preprocessing on the work order structure data and the work order unstructured data to obtain an entity-relation-attribute chain; constructing an initial knowledge graph according to the entity-relation-attribute chain, and then serving the initial knowledge graph and connecting the initial knowledge graph with a work order management system; integrating work order initiation channels by using the work order management system, synchronously collecting electronic work order source data in real time, then classifying the electronic work order source data according to the entity-relation in the initial knowledge graph, and performing optimal resource matching on the classified electronic work order source data to obtain electronic work order standard source data, which is used to optimize the initial knowledge graph for the first time to obtain a standard knowledge graph; generating a preferred solution list and a knowledge push list dynamically according to the standard knowledge graph and in combination with the real-time collection of electronic work order source data, and then recycling work order processing process information; performing work order archiving according to the work order processing process information, optimizing the standard knowledge graph for the second time to obtain an optimized standard knowledge graph, and then connecting the work order management system with the optimized standard knowledge graph; the method for defining core entities of the work order management field, thereby defining relationships between the core entities, obtaining a core entity relationship chain, and then determining attributes corresponding to each core entity to construct a work order management field ontology graph, specifically comprises the following steps: based on the business scenario requirements of the work order management field, carrying out field research to obtain business process documents, historical work order records and field expert experience data related to the work order management; screening out an entity candidate set in which the frequency of appearance of entities in historical work order records is greater than or equal to a preset threshold value through the business process documents, the historical work order records and the field expert experience data; based on the entity candidate set, inviting field experts to score the core degrees of the candidate entities, screening out entities with core degree scores greater than or equal to 3 to obtain a core entity list of the work order management field; analyzing the interaction scenarios of each core entity in the business process, counting the association frequencies between entities, and determining the relationship weights between entities; screening out entity relationships with relationship weights greater than or equal to a preset relationship weight threshold value, and sorting the entity relationships in descending order of association frequencies to obtain a core entity relationship chain; based on the core entity relationship chain and the business scenario requirements of the core entities, determining attribute items corresponding to each core entity, and simultaneously determining the integrity requirement coefficients of each attribute item, and screening out attribute items with integrity requirement coefficients greater than or equal to 0.9 as core attributes; based on the core entity list, the core entity relationship chain and the core attributes of the core entities, defining semantic constraints of entities, relationships and attributes, and generating a work order management field ontology graph. 2.The knowledge graph-based work order full-process electronic management method according to claim 1, characterized in that, The work order management domain ontology graph is based on the collection of work order data sources, which are divided into structured and unstructured according to the synchronization, and the work order structure data and the work order unstructured data are obtained, and then the data preprocessing is carried out on the work order structure data and the work order unstructured data respectively, so that the entity-relation-attribute chain is obtained, which specifically includes: Based on the work order management domain ontology graph, the collection range of the work order data source is determined, the collection range includes historical paper work order scanning, historical electronic work order in the work order management system, business data of the business association system, processing manual and training document written by the domain expert, and the original work order data source set is formed; Based on the original work order data source set, the data format characteristics of each data source are extracted, and then the matching degree of the data source and the ontology graph core field is determined; Based on the matching degree and the data format characteristics, the original work order data source set is divided into work order structure data and work order unstructured data; The redundant data and abnormal data of the work order structure data are removed to obtain the cleaned structured data; Based on the cleaned structured data, the data standardization processing is carried out according to the attribute definition of the ontology graph, the same attribute value in different formats is unified into a preset format, and the standardized structured data is obtained; The text processing is carried out on the work order unstructured data to obtain the text unstructured data; The text unstructured data is matched with the core entity in the work order management domain ontology graph to obtain the entity matching result; Based on the entity matching result, the entity-relation pair in the unstructured data is determined combined with the core entity relationship chain in the work order management domain ontology graph, and the attribute information corresponding to the entity is extracted to obtain the entity-relation-attribute set of the unstructured data; Based on the standardized structured data, the core entity, relationship and attribute in the ontology graph are directly mapped to obtain the entity-relation-attribute set of the structured data; The entity-relation-attribute set of the structured data and the entity-relation-attribute set of the unstructured data are fused to obtain the entity-relation-attribute chain. 3.The knowledge graph-based work order full-process electronic management method according to claim 1, characterized in that, According to the entity-relation-attribute chain, the initial knowledge graph is constructed, and then the initial knowledge graph is serviced and connected with the work order management system, which specifically includes: According to the entity-relation-attribute chain, the number of entities, the number of relationships and the number of attributes in the chain are determined to obtain the storage demand size of the graph; Based on the storage demand size, the adaptive graph database is determined, and then the knowledge graph template is created, the entity label, the relationship type and the attribute key are defined to ensure the consistency of the structure with the entity-relation-attribute chain; Based on the created knowledge graph template, the data in the entity-relation-attribute chain is imported into the graph database in batches; After all the data are imported, the entity index and the relationship index are constructed, the index fields include entity ID, core attribute, the query response time after testing the index construction is determined to obtain the initial knowledge graph; Based on the initial knowledge graph, the graph service API is developed; Based on the graph service API, the API test case is written, and the API test case is executed to complete the service of the initial knowledge graph; The interface document of the work order management system is obtained, the interface type, request protocol and data interaction format of the system are analyzed; Based on the interface information of the work order management system, a docking adapter of the atlas service and the work order management system is developed to realize data format conversion and protocol adaptation between the two; After the docking is completed, joint debugging test is performed to simulate the scene of the work order management system calling the atlas service API, and the initial knowledge graph is docked with the work order management system. 4.The knowledge graph-based work order full-process electronic management method according to claim 1, characterized in that, The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the work order management system. The initial knowledge graph is docked with the work order management system, and the initial knowledge graph is docked with the 5.The knowledge graph-based work order full-process electronic management method according to claim 1, characterized in that, The preferred solution list and the knowledge push list are dynamically generated according to the standard knowledge graph combined with real-time collection of electronic work order source data, and then the work order processing flow information is recovered, specifically including: According to the standard knowledge graph, the knowledge asset data related to the work order processing is extracted, including the historical solution set, the associated SOP document set, and the historical case set; Based on the real-time collection of electronic work order source data, the key features of the current work order are extracted, including the work order type, priority, fault description, and associated entities; The similarity of the key features of the current work order to the features of each solution in the historical solution set is obtained, and the solution priority function is defined by combining the historical performance indicators of the solution, so as to determine the solution priority value; The historical solutions are sorted in descending order of the solution priority value, and the solutions are screened to form the preferred solution list; From the associated SOP document set of the standard knowledge graph, the SOP documents matching the key features of the current work order are filtered to obtain the SOP recommendation set; From the historical case set, the cases with a similarity greater than or equal to a preset similarity threshold to the key features of the current work order are filtered to obtain the case recommendation set; The SOP recommendation set and the case recommendation set are arranged in the order of SOP document priority and case second, to form the knowledge push list; The preferred solution list and the knowledge push list are pushed to the processing interface of the work order management system for the processing personnel to view and select; The processing personnel carry out work order processing work based on the preferred solution list and the knowledge push list, and the work order management system records the key information in the processing process in real time, including the processing start time, processing step record, resources called in the processing process, and processing end time; When the work order processing state changes to processing completion, the work order management system automatically generates a processing result questionnaire and pushes it to the work order initiator user, and sets a questionnaire recovery time limit; If user feedback is received within the questionnaire recovery time limit, the feedback information is integrated with the processing process record; If no feedback is received beyond the questionnaire recovery time limit, the problem is marked as solved by default, and a note is recorded for the user's non-feedback; Based on the integrated processing process record and user feedback information, complete work order processing flow information is formed. 6.The knowledge graph-based work order full-process electronic management method according to claim 1, characterized in that, According to the work order processing flow information, the work order is archived, and the standard knowledge graph is optimized for the second time to obtain the preferred standard knowledge graph, specifically including: According to the work order processing flow information, the core identification information of the work order is extracted, including the work order ID, processing state, processing resource ID, and user feedback result, as the core basis for archiving classification; Based on the core identification information, the work order is classified and archived, and the corresponding work order processing flow information and associated electronic work order standard source data are stored under each classification directory; The number of work orders in each archived classification directory is obtained, and if the number of work orders in the archived classification directory is greater than or equal to a preset compression threshold, the files under the directory are compressed; If the number of work orders in the archived classification directory is less than the preset compression threshold, the compression algorithm is replaced and re-compressed until the number of work orders in the archived classification directory is greater than or equal to the preset compression threshold, and the work order archiving is completed. Based on the archived work order processing flow information, the actual application effect of each entity-relation-attribute combination in the standard knowledge graph is counted, including the frequency of entity invocation, the actual matching accuracy of the relationship, and the effectiveness coefficient of the attribute. If there is an entity whose invocation frequency is less than or equal to a preset low frequency threshold, it is marked as a low frequency entity, and it is checked whether it is a newly added entity. If it is a newly added entity, it is retained. If it is a redundant entity, it is deleted from the standard knowledge graph. If there is a relationship with an actual matching accuracy less than a preset accuracy threshold, the reason for the relationship matching error is analyzed, and the definition or matching rule of the relationship is corrected. If there is an attribute with an effectiveness coefficient less than a preset effectiveness threshold, it is determined to be a low-value attribute, and it is evaluated whether it is a core attribute. If it is not a core attribute, it is removed from the attribute list of the corresponding entity. Based on the results of low-frequency entity processing, relationship correction, and attribute optimization, the entity set, relationship set, and attribute set of the standard knowledge graph are updated, and new entities and new relationships appearing in the archived work order are supplemented to the standard knowledge graph to obtain an optimized graph. The graph accuracy of the optimized graph is obtained. If the graph accuracy of the optimized graph is less than a preset graph accuracy threshold, the archived work order data is reanalyzed to find the incorrect entity-relation-attribute combination and correct it until the graph accuracy of the optimized graph is greater than or equal to the preset graph accuracy threshold. At the same time, the knowledge coverage of the optimized graph is obtained. If the knowledge coverage of the optimized graph is less than a preset knowledge coverage threshold, entities, relationships, and attributes that are not covered by the business scenario are supplemented until the knowledge coverage of the optimized graph is greater than or equal to the preset knowledge coverage threshold to obtain an optimized standard knowledge graph.

7. A knowledge graph-based work order full-process electronic management system for implementing the management method of any one of claims 1-6, characterized in that, It includes: A work order domain ontology construction module, which is used to define core entities in the work order management domain, thereby defining the relationships between the core entities, obtaining a core entity relationship chain, and then determining the attributes corresponding to each core entity to construct a work order management domain ontology graph; A work order data collection and preprocessing module, which is used to collect work order data sources based on the work order management domain ontology graph, and synchronously divide the work order data sources into structured and unstructured to obtain work order structured data and work order unstructured data, and then preprocess the work order structured data and work order unstructured data respectively to obtain entity-relation-attribute chains; An initial knowledge graph construction and system interface module, which is used to construct an initial knowledge graph according to the entity-relation-attribute chains, and then service the initial knowledge graph and interface it with the work order management system. The work order source data processing and knowledge graph first optimization module is configured to integrate work order initiation channels by using a work order management system, synchronously collect electronic work order source data in real time, then classify the electronic work order source data according to entities-relationships in an initial knowledge graph, perform optimal resource matching on the classified electronic work order source data, obtain electronic work order standard source data, and use the electronic work order standard source data to perform first optimization on the initial knowledge graph, thereby obtaining a standard knowledge graph. The work order knowledge application and knowledge graph second optimization module is configured to dynamically generate an optimal solution list and a knowledge push list according to the standard knowledge graph and in combination with real-time collection of electronic work order source data, and then recycle work order processing flow information; the work order knowledge application and knowledge graph second optimization module is also configured to perform work order archiving according to the work order processing flow information, perform second optimization on the standard knowledge graph, thereby obtaining an optimal standard knowledge graph, and then connect the work order management system with the optimal standard knowledge graph. 8.The knowledge graph-based work order full-process electronic management system according to claim 7, characterized in that, The work order domain ontology construction module includes: A core entity and relationship definition unit is configured to define core entities in the work order management domain, thereby defining relationships between the core entities, and obtaining a core entity relationship chain. A core entity attribute determination unit is configured to determine attributes corresponding to each core entity after the core entity relationship chain is obtained. A work order domain ontology graph construction unit is configured to construct a work order management domain ontology graph based on the core entity relationship chain and the attributes corresponding to each core entity. 9.The knowledge graph-based work order full-process electronic management system according to claim 7, characterized in that, The work order data collection and preprocessing module includes: A work order data source collection unit is configured to collect work order data sources based on the work order management domain ontology graph. A work order data classification unit is configured to synchronously divide the collected work order data sources into structured and unstructured data, thereby obtaining work order structured data and work order unstructured data. A work order data preprocessing unit is configured to perform data preprocessing on the work order structured data and the work order unstructured data, thereby obtaining an entity-relationship-attribute chain.

Images