Industrial intelligent agent-oriented knowledge center driving method and system

Abstract

This invention discloses a knowledge center-driven method and system for industrial intelligent agents, belonging to the fields of industrial software, artificial intelligence, knowledge engineering, and intelligent agent workflow technology. The method establishes a three-layer linkage architecture: a backend knowledge center, a business intelligent agent workflow, and a frontend interaction layer. The backend receives multi-source heterogeneous industrial data, parses it modally to generate knowledge objects, constructs a constraint- and version-based pattern layer, and builds three types of knowledge bases: graph, vector, and form. Six major knowledge capabilities are encapsulated as configurable nodes. Frontend business requirements are transformed into temporary knowledge objects and trigger workflows. Valid temporary knowledge is automatically converted into active knowledge, and interactive data is written back to the knowledge base to complete iteration. The three-layer architecture forms a knowledge loop through a unified identifier. This invention achieves unified modeling, modular scheduling, and autonomous accumulation and updating of industrial knowledge, improving knowledge reuse, process configuration, and asset iteration capabilities. It is applicable to industrial intelligent design scenarios such as ship foundations.

Description

Technical Field

[0001] This invention belongs to the fields of industrial software, artificial intelligence, knowledge engineering and intelligent agent workflow technology, and specifically relates to a knowledge center-driven method and system for industrial intelligent agents. Background Technology

[0002] In the process of industrial digitalization and intelligent transformation and upgrading, industrial AI agents have been widely applied in complex industrial business scenarios such as product forward design, process planning, automatic solution generation, quality defect analysis, equipment lifecycle operation and maintenance, and domain knowledge management. Compared with traditional closed industrial software, industrial agents have dynamic task processing capabilities, can respond to natural language requirements, parse business drawings / 3D model files, and link knowledge bases, knowledge graphs, and large language models to complete tasks such as knowledge retrieval, intelligent parameter recommendation, industry rule verification, structural logic reasoning, and automatic generation of customized solutions.

[0003] However, existing technologies have the following shortcomings: The existing knowledge center can receive multimodal industrial data and complete basic storage, but the complete chain of multi-source heterogeneous data—from import, parsing, schema-level modeling, knowledge graph construction to providing knowledge services to intelligent agents—cannot be established. After parsing, heterogeneous data such as text, spreadsheets, engineering drawings, 3D models, industry standard documents, historical business cases, and expert experience are stored independently in vector libraries, structured forms, and file index libraries. There is no unified, standardized knowledge object association system for these various knowledge assets. Industry rules and historical business cases are stored only as independent documents and isolated data tables, unable to be transformed into searchable rule nodes, case nodes, and entity relationships within the knowledge graph. This prevents industrial intelligent agents from automatically matching knowledge types and selecting retrieval strategies based on task context, and also from automatically updating the knowledge base with temporary business data. Knowledge retrieval and iteration lack a unified execution standard.

[0004] In existing technologies, underlying capabilities such as knowledge graph retrieval, industry rule query, intelligent parameter recommendation, and knowledge logic reasoning are fixed as program code or independent plugins, and cannot be broken down into standardized, configurable functional nodes embedded in the workflows of different business intelligent agents. For different industrial business scenarios and different task stages, technical personnel need to repeatedly develop and configure retrieval, recommendation, and reasoning logic, resulting in high development costs; moreover, it is impossible to combine multiple types of knowledge capabilities for a single task stage, or to reuse the same knowledge capability across multiple task stages, making it difficult to adapt to differentiated industrial business needs.

[0005] Temporary files, 3D models, and natural language requests uploaded during user front-end interactions only serve a single session or tool node, and cannot be integrated into the standardized knowledge object system of the backend, thus preventing their reuse in subsequent workflow tasks. When receiving knowledge entry instructions from users, effective business knowledge, rule verification results, and new business cases with reusable value cannot be automatically written back to the formal knowledge base. Incremental knowledge generated by front-end interactions cannot be standardized and accumulated, resulting in permanent breakpoints in the front-end and back-end data links.

[0006] Therefore, the existing knowledge architecture supporting industrial intelligent agents suffers from technical bottlenecks due to the disconnect between knowledge storage, workflow configuration, and front-end interaction. The industry urgently needs a knowledge center-driven solution for industrial intelligent agents that can connect the entire chain of back-end data processing, knowledge asset organization, knowledge capability node release, business workflow configuration, front-end interaction, and knowledge iteration. Based on a graph-based structure to unify management rules and case-based industrial knowledge, it can build standardized knowledge assets that are searchable, inferable, and dynamically updated, providing unified knowledge services and a knowledge accumulation carrier for business intelligent agents. Summary of the Invention

[0007] Existing technologies suffer from several drawbacks: fragmented backend knowledge centers, business-side intelligent agent workflows, and frontend application interaction modules; low industrial knowledge reuse rates; insufficient flexibility in business workflow configuration; and the inability to continuously iterate and evolve knowledge assets. The purpose of this invention is to provide a knowledge center-driven method and system for industrial intelligent agents. This method achieves the graphing, modularization, reusability, and sustainable iteration of industrial knowledge assets through multi-source heterogeneous industrial data knowledge processing, knowledge capability node encapsulation, flexible intelligent agent workflow configuration, frontend human-computer interaction, and automatic knowledge feedback and updating across the entire chain.

[0008] To achieve the above objectives, the present invention provides a knowledge center-driven method and system for industrial intelligent agents, comprising the following steps: S1. Receive multi-source heterogeneous data from industrial operations in the backend knowledge center; the multi-source heterogeneous data includes at least one of text, tables, drawings, models, rule documents, historical cases, and expert experience data; S2. Select the corresponding parsing strategy according to the data mode of the multi-source heterogeneous data, and perform parsing extraction, structured transformation, vectorization representation and rule-based structured processing on the multi-source heterogeneous data to form data objects and knowledge objects; S3. Construct a schema layer object based on the knowledge object; the schema layer object includes entity type, relation type, attribute field, field mapping, rule constraint, associated file and version identifier; organize and store knowledge assets according to the schema layer object to form a knowledge graph library, vector library and structured form library; wherein, rule knowledge and case knowledge are stored, retrieved and called in the form of graph nodes, relation edges, attributes and subgraphs; S4. Encapsulate the graph retrieval capability, vector retrieval capability, rule query capability, knowledge recommendation capability, knowledge reasoning capability, and knowledge update and accumulation capability corresponding to the knowledge asset into a knowledge capability node; during the configuration process of the business-side intelligent agent workflow, receive the selection and configuration of the knowledge capability node, and embed the knowledge capability node into one or more task stages in the business scenario intelligent agent workflow; the knowledge capability node can be configured to multiple task stages in the same business scenario intelligent agent workflow, and the same task stage can call multiple knowledge capability nodes simultaneously. S5. During the front-end application interaction process, receive business requirements submitted by users through file upload, model submission, or language interaction, process the business requirements into temporary knowledge objects, call the business scenario intelligent agent workflow, and generate business execution results; when the temporary knowledge object is called by the business scenario intelligent agent workflow and confirmed by the user to have reusable value, upgrade the temporary knowledge object into a formal knowledge object. S6. Write back the business execution results, user feedback, temporary knowledge objects, operation records and new cases to the backend knowledge center to update the knowledge graph library, vector library, structured form library, pattern layer object or knowledge capability node configuration; The backend knowledge center, the business-side intelligent agent workflow, and the frontend application interaction are globally linked through knowledge object identifiers, pattern layer versions, knowledge capability node identifiers, business task identifiers, and operation record identifiers.

[0009] Preferably, in S2, the data object and knowledge object include: The file object is used to record the file name, file type, file source, uploading user, parsing status, and associated business scenario; The parsed result object is used to record text slices, table fields, drawing areas, model components, rule clauses, and case fields; Vector knowledge objects are used to record vector indexes, semantic representations, and similarity retrieval entry points; Graph knowledge objects are used to record entities, relationships, attributes, constraints, paths between entities, rule nodes, case nodes, rule triggering relationships, case similarity relationships, and subgraph structures; A structured form object is used to record fields, parameters, headers, cells, and structured records, and to establish mapping relationships with entities, attributes, or rule nodes in the graph knowledge object.

[0010] Preferably, in S3, the construction of the schema layer object based on the knowledge object specifically includes: The uploaded business data is parsed and segmented to form searchable text, intermediate structure data, and candidate knowledge objects; By combining vector retrieval results, existing references, and user input, entities, relationships, attributes, and rule constraints are constructed in the schema layer canvas; Candidate entities, candidate relationships, candidate attributes, and candidate rules are generated based on associated files; It receives user confirmation, modification, or deletion operations for candidate content and outputs a schema layer object with a version identifier.

[0011] Preferably, in S4, the knowledge capability node includes: The graph retrieval node is used to retrieve entities, relationships, paths, rule nodes, case nodes, subgraphs, and contextual knowledge from the knowledge graph database. Vector retrieval nodes are used to retrieve documents, cases, or drawing fragments that are similar to user requirements, design documents, or model descriptions from the vector library; The rule query node is used to query design criteria, verification rules, or process constraints applicable to the current task stage based on rule nodes, rule relationships, and rule application conditions in the knowledge graph database. Knowledge recommendation nodes are used to generate business suggestions, candidate solutions, or parameter suggestions based on user needs, historical cases, rule nodes, and graph relationships. Knowledge reasoning nodes are used to generate reasoning conclusions based on rule graph relationships, case graph relationships, parameter constraints, structural relationships, or process states. Knowledge update accumulation nodes are used to accumulate newly added knowledge, operation records, business results or case results after user confirmation to the backend knowledge center, and update the corresponding graph nodes, relationships, attributes, subgraphs or vector indexes.

[0012] This invention also provides a knowledge center-driven system for industrial intelligent agents, comprising: The backend knowledge center module is used to execute S1, S2, and S3, receive multi-source heterogeneous data, and complete data parsing, schema layer construction, knowledge asset organization, and knowledge base updates. The knowledge asset management module is used to execute S3, manage the knowledge graph library, vector library, and structured form library, and manage rule knowledge and case knowledge in the form of graph nodes, relation edges, attributes, and subgraphs; The knowledge capability node management module is used to execute S4, encapsulating graph retrieval, vector retrieval, rule query, knowledge recommendation, knowledge reasoning, and knowledge update accumulation into configurable nodes; The business-side intelligent agent workflow configuration module is used to execute S4 to configure the knowledge capability nodes to one or more task stages in the business scenario intelligent agent workflow. The front-end application interaction module is used to execute S5, receive user-uploaded files, submit models or language interaction input, and display the business execution results; The knowledge update module is used to execute S6 and update the backend knowledge center based on user feedback, operation records, temporary knowledge objects, and new cases.

[0013] Therefore, the present invention employs the above-mentioned knowledge center-driven method and system for industrial intelligent agents. Compared with the prior art, the technical solution of the present invention has the following beneficial effects: (1) This invention builds a three-layer collaborative linkage architecture, analyzes industrial multi-source data according to data modality differences, constructs a pattern layer with constraints and version identifiers, and stores rules and cases in a unified graph form to establish a standardized knowledge object system; it solves the defects of scattered and fragmented multi-source data, fragmented case documents, broken data links in the three-layer architecture, and lack of unified modeling specifications for knowledge, which makes it impossible for intelligent agents to accurately match and call appropriate knowledge, realizes integrated knowledge modeling of industrial data, opens up the entire data processing link, unifies and standardizes knowledge storage and retrieval logic, and effectively improves the level of knowledge structuring and retrieval efficiency; (2) This invention encapsulates six categories of knowledge capabilities into configurable independent nodes, supports the reuse of knowledge nodes across task stages and parallel scheduling of multiple nodes in a single stage, and can be flexibly embedded into any link of the intelligent agent workflow; it overcomes the shortcomings of traditional solutions where knowledge capabilities are fixed, difficult to split and reuse, and knowledge calling logic needs to be repeatedly developed for each business scenario, and cannot adapt to the differentiated knowledge needs of different task stages, realizes modular and flexible configuration of knowledge services, saves the work of repeated development, adapts to the needs of multi-stage business, reduces the development cost of intelligent agent process, and improves the flexibility of workflow configuration; (3) This invention transforms front-end business requirements into standardized temporary knowledge objects, and business interaction data is uniformly written back to the knowledge base. Temporary knowledge that has been confirmed by users to have reuse value is automatically converted into formal assets. Based on the unique identifier, the three-layer architecture is connected to form a knowledge iteration closed loop. It solves the problems that front-end temporary data only takes effect once, new business cases cannot be stored in the database, there are data breakpoints between the front and back ends, and the knowledge base is difficult to update autonomously. It builds a knowledge self-accumulation and self-updating closed loop system, continuously expands the knowledge base reserves, and significantly improves the accuracy of industrial intelligent body solution generation and parameter recommendation.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a flowchart of an embodiment of a knowledge center-driven method and system for industrial intelligent agents according to the present invention; Figure 2 This is a schematic diagram of the three-layer linkage architecture and full-link data flow of a knowledge center-driven method and system for industrial intelligent agents according to the present invention (Example 2). Figure 3This is a schematic diagram of the knowledge-driven collaborative closed-loop mechanism in Embodiment 3 of the knowledge center-driven method and system for industrial intelligent agents according to the present invention. Figure 4 This is a schematic diagram of a ship pedestal intelligent design scenario, which is the fourth embodiment of the knowledge center-driven method and system for industrial intelligent agents according to the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used in the present invention should have the ordinary meaning understood by those skilled in the art.

[0017] Example 1 like Figure 1 As shown, the knowledge center-driven method for industrial intelligent agents according to the present invention includes the following steps: S1. The backend knowledge center uniformly accesses various types of data from industrial business scenarios, including text specifications, spreadsheets, engineering drawings, 3D models, industry rule documents, historical business cases, and expert experience; long-term reused data is stored in the formal knowledge base, while data temporarily uploaded for a single task is stored in an independent temporary knowledge partition. S2. The system matches exclusive parsing strategies based on data modalities: for text and rule documents, it performs text slicing, clause recognition, rule condition extraction, and semantic vectorization representation; for spreadsheets, it performs field recognition, header standardization, and structured form generation; for engineering drawings and 3D models, it performs drawing area segmentation, model component indexing, and geometric metadata extraction; for historical business cases, it performs case field extraction, scheme element recognition, and case relationship extraction.

[0018] After parsing, file objects, parsing result objects, vector knowledge objects, graph knowledge objects, and structured form objects are uniformly generated. Industry rules and historical cases are not stored separately but are uniformly transformed into graph rule nodes, case nodes, applicable associations, triggering associations, similar associations, and independent subgraphs, which can be called through various knowledge capability nodes. After objectification and graphification, multi-source heterogeneous original files are transformed into knowledge base assets that can be called in a standardized manner by intelligent agent workflows. S3. Construct a schema layer object based on the parsed standardized knowledge object; the schema layer object is used to define the knowledge organization specifications, including entity type, relationship type, attribute fields, field mapping rules, business constraint rules, associated attachment files, and version unique identifier.

[0019] Under the constraints of the schema layer objects, the system generates three types of knowledge assets: a knowledge graph library, a vector library, and a structured form library. The knowledge graph library stores entities, entity relationships, node attributes, entity paths, rule nodes, case nodes, and independent subgraphs. The vector library is used for semantic matching and similar case retrieval. The structured form library stores business parameter tables and field record tables, and establishes mapping associations with graph entities, node attributes, and rule nodes. S4. After the backend knowledge center completes the standardized organization of knowledge assets, it encapsulates the functions of graph retrieval, vector retrieval, rule query, knowledge recommendation, knowledge reasoning, and knowledge update and accumulation into independent standardized knowledge capability nodes. Each knowledge capability node is bound to a unique node identifier, input object specifications, output object specifications, associated knowledge base, adapted task stage, call triggering conditions, and standardized return format.

[0020] This step transforms the backend underlying knowledge service interface into a standardized functional module that can be directly dragged and configured in the business workflow, eliminating the coupling of the underlying code; S5. Business personnel select the corresponding standardized knowledge and capability nodes according to the target business scenario and task stage; a single knowledge and capability node can be reused in multiple task stages, and a single task stage can combine and call multiple types of knowledge and capability nodes.

[0021] Example configuration: During the demand identification phase, vector retrieval nodes and graph retrieval nodes are called in parallel; during the parameter generation phase, rule query nodes and knowledge recommendation nodes are called; during the rule verification phase, rule query nodes and knowledge reasoning nodes are called; during the result output and feedback update phase, knowledge update and accumulation nodes are called.

[0022] Front-end user-uploaded files, 3D models, and natural language requirements are uniformly converted into temporary knowledge objects, bound to corresponding knowledge bases, temporary knowledge partitions, and workflow task nodes, and automatically triggered intelligent agent workflow execution. S6. The system will uniformly transmit the business execution results of the intelligent agent workflow, user interaction feedback, system operation records, temporary knowledge objects, and newly added business cases back to the backend knowledge center. If the user confirms that the temporary knowledge object has reusable value, it will be upgraded to a formal standardized knowledge object, and the knowledge graph library, vector library, structured form library, case independent subgraph, pattern layer object, and knowledge capability node configuration parameters will be updated synchronously.

[0023] This step forms a closed loop of the entire chain: "knowledge import - knowledge modeling - node encapsulation - business call - front-end interaction - knowledge feedback and update".

[0024] Example 2 like Figure 2As shown, the knowledge center-driven system for industrial intelligent agents of this invention is divided into a three-layer linkage architecture, including a backend knowledge center layer, a business-side intelligent agent workflow layer, and a frontend application interaction layer. The three-layer architecture establishes a global association index through unique identifiers of knowledge objects, pattern layer version numbers, knowledge capability node identifiers, business task numbers, and unique identifiers of execution records. The specific functions and interaction relationships of the three layers are as follows: The backend knowledge center layer corresponds to the backend knowledge center module, knowledge asset management module, and knowledge capability node management module in the system claims. It undertakes the functions of multimodal data access, standardized parsing, pattern layer object construction, knowledge asset storage, and knowledge capability node encapsulation. Within the layer, it generates all knowledge assets, including pattern layer objects, knowledge graph library, vector library, and structured form library. Through a standardized and unified interface, it outputs configurable standardized knowledge capability nodes to provide underlying knowledge service support for the intelligent agent workflow on the business side.

[0025] The business-side intelligent agent workflow layer corresponds to the business-side intelligent agent workflow configuration module in the system claims, including a business scenario definition unit, a multi-level task stage unit, a standardized knowledge and capability node scheduling unit, and a workflow execution unit. Standardized knowledge and capability nodes have no fixed task stage binding relationship and can be reused across task stages according to business needs: graph retrieval nodes can adapt to the conceptual solution stage's model matching and the geometric structure generation stage's structural relationship reasoning; rule query nodes, relying on graph rule nodes and applicable associations, can synchronously adapt to the parameter generation stage and the result verification stage; vector retrieval nodes are used for similar historical case retrieval and design document semantic recall; knowledge update and accumulation nodes can be configured in the result output stage and the user feedback knowledge update stage. The workflow execution unit receives business requirement instructions from the front end, schedules the corresponding knowledge and capability nodes to complete business calculations, and outputs standardized business execution results.

[0026] The front-end application interaction layer corresponds to the front-end application interaction module and knowledge update module in the system claims. It receives three types of input commands from users: file upload, 3D model submission, and natural language interaction. It converts the input content into temporary knowledge objects and sends them to the business-side intelligent agent workflow. The business execution results output by the workflow are visualized at this layer, while collecting user confirmation feedback on the solution and the newly added temporary knowledge. The incremental knowledge and interaction records after user confirmation are sent back to the back-end knowledge center layer through the backflow link to complete the incremental update of the entire domain knowledge assets.

[0027] The three-tier architecture enables data interoperability through a unified identifier system, eliminating independent data silos and connecting the entire chain of knowledge storage, knowledge retrieval, and knowledge updates.

[0028] Example 3 like Figure 3As shown, this embodiment illustrates the closed-loop operation logic of the entire data cyclic iteration process of the present invention. The closed-loop operation mechanism of the present invention includes four core execution stages that are interconnected, with data flowing unidirectionally from one stage to the other, and the data flowing back from the end stage to the beginning stage, forming a continuously iterative closed loop: The multimodal knowledge import process receives heterogeneous industrial raw data from multiple sources, including text, spreadsheets, engineering drawings, 3D models, industry rule documents, and historical cases; it outputs the raw data to the data parsing and pattern modeling units as the raw material for knowledge base construction.

[0029] The knowledge base standardization construction phase receives the raw data output from the import phase, performs data parsing and schema layer object construction operations, and generates schema layer objects, knowledge graph library, vector library, and structured form library; industry rule knowledge and historical business cases are stored in the form of graph nodes, entity relationship edges, and independent subgraphs; at the end of the phase, various knowledge service capabilities are encapsulated into standardized knowledge capability nodes and published externally, and configurable knowledge capability nodes are output to the business workflow configuration phase.

[0030] The business workflow node configuration stage receives standardized knowledge and capability nodes and completes the binding configuration of nodes and task stages according to the business scenario; after receiving the business call instructions issued by the front end, it schedules the corresponding knowledge and capability nodes to perform calculations and outputs standardized business execution results to the user application interaction stage.

[0031] The user application interaction stage receives the execution results output by the business workflow and displays them to the user in a visual format. At the same time, it collects the business requirements input by the user, user confirmation feedback, and incremental temporary knowledge objects. New cases, interaction records, and valid temporary knowledge after user confirmation are fed back to the multimodal knowledge import stage and the knowledge base standardization construction stage to complete the iterative update of knowledge assets.

[0032] The four key processes operate in a continuous cycle, enabling the continuous accumulation of industrial knowledge, the continuous optimization of knowledge service capabilities, and the continuous iteration of business design solutions.

[0033] Example 4 like Figure 4 As shown, this embodiment is a specific industrial application example of a general solution. It fully verifies the feasibility of the technical solution of this invention using a sub-scenario of intelligent ship foundation design, and fully discloses the entire process of data processing, workflow configuration, and knowledge feedback logic in that scenario.

[0034] The backend knowledge center accesses a dedicated multi-source data source for the ship foundation field, which includes STEP / SLDPRT format 3D models of foundations, foundation design specification documents, engineering assembly drawings, equipment parameter spreadsheets, and historical ship foundation design cases.

[0035] The system performs modality matching parsing, structure transformation, and vectorization representation processing on all data sources to generate three types of domain-specific knowledge assets: Knowledge graph library: Stores base component entities, structural relationships, component parameter constraints, rule graph nodes, and independent subgraphs of historical cases; strength, material, and process clauses in the base design specifications are converted into rule nodes, and applicable condition associations and trigger associations are configured; historical base cases are encapsulated as independent case subgraphs, and similarity association edges are established for cases; Vector library: Stores semantic vector representations of base design documents, drawings, and historical cases for quick retrieval of similar solutions; Structured form library: Stores ship equipment parameters, base component size parameters, and strength verification calculation results. A one-to-one mapping relationship is established between form fields and entities, attributes, and rule nodes in the map.

[0036] On the business side, a three-stage intelligent agent workflow for ship foundation design is built, with standardized knowledge and capability nodes configured as needed for each task stage. Nodes support cross-stage reuse. Conceptual design phase: Configure graph retrieval nodes and vector retrieval nodes; graph retrieval nodes complete historical base model matching based on equipment model, installation conditions, and load constraints; vector retrieval nodes recall semantically similar historical complete base design cases; Parameter design phase: Configure rule query nodes, knowledge recommendation nodes, and vector retrieval nodes; rule query nodes retrieve relevant industry constraint rules for plate thickness, hole edge distance, weld seam, and reinforcing components within the atlas; knowledge recommendation nodes output recommended parameters for components based on atlas entity associations and historical case sub-atlases; Geometric design phase: Configure knowledge reasoning nodes, graph retrieval nodes, and rule query nodes; the knowledge reasoning nodes generate the base geometry based on the topological relationships of the panel, web, elbow, and hull structure; the rule query nodes perform compliance verification on the generated geometry.

[0037] The knowledge update and accumulation nodes are uniformly configured in the two sub-stages of the workflow end result output and user feedback confirmation.

[0038] Front-end users upload a ship foundation design requirements document and an initial 3D reference model of the foundation. The requirements document records the equipment weight, installation space, load conditions, hull structural constraints, and design objectives. The 3D model includes a reference foundation and local structural models. The system converts the uploaded content into temporary knowledge objects specific to the ship foundation scenario, binds them to the foundation domain knowledge base and foundation design workflow, and automatically starts the intelligent agent workflow execution.

[0039] The workflow outputs corresponding business results in stages: the concept stage outputs candidate parent schemes, the parameter stage outputs component size parameters, and the geometry stage outputs complete base structure schemes; finally, an independent sub-map of the target base is generated, which includes base component nodes, structural association edges, component parameter nodes, 3D model link nodes, constraint rule nodes, and design result record nodes.

[0040] After the user completes the review of the target foundation scheme and confirms its reusability, the system writes back the target foundation sub-graph, 3D model links, component parameters, verification trigger rules, and complete design records corresponding to this design as newly added formal standardized knowledge objects to the backend knowledge center; simultaneously, it updates the ship foundation domain knowledge graph library, vector library, and structured form library. Subsequent ship foundation design tasks can directly access the accumulated case knowledge to complete scheme matching and parameter recommendations.

[0041] Therefore, the present invention adopts the above-mentioned knowledge center-driven method and system for industrial intelligent agents. This method realizes the graphing, modularization, reusability, and sustainable iteration of industrial knowledge assets through multi-source heterogeneous industrial data knowledge processing, knowledge capability node encapsulation, flexible configuration of intelligent agent workflow, front-end human-computer interaction, and automatic knowledge feedback and updating throughout the entire chain.

[0042] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A knowledge-center driven method for industrial intelligent agents, characterized in that, Includes the following steps: S1. Receive multi-source heterogeneous data from industrial operations in the backend knowledge center; the multi-source heterogeneous data includes at least one of text, tables, drawings, models, rule documents, historical cases, and expert experience data; S2. Select the corresponding parsing strategy according to the data mode of the multi-source heterogeneous data, and perform parsing extraction, structured transformation, vectorization representation and rule-based structured processing on the multi-source heterogeneous data to form data objects and knowledge objects; S3. Construct a schema layer object based on the knowledge object; the schema layer object includes entity type, relation type, attribute field, field mapping, rule constraint, associated file and version identifier; organize and store knowledge assets according to the schema layer object to form a knowledge graph library, vector library and structured form library; wherein, rule knowledge and case knowledge are stored, retrieved and called in the form of graph nodes, relation edges, attributes and subgraphs; S4. Encapsulate the graph retrieval capability, vector retrieval capability, rule query capability, knowledge recommendation capability, knowledge reasoning capability, and knowledge update and accumulation capability corresponding to the knowledge asset into a knowledge capability node; during the configuration process of the business-side intelligent agent workflow, receive the selection and configuration of the knowledge capability node, and embed the knowledge capability node into one or more task stages in the business scenario intelligent agent workflow; the knowledge capability node can be configured to multiple task stages in the same business scenario intelligent agent workflow, and the same task stage can call multiple knowledge capability nodes simultaneously. S5. During the front-end application interaction process, receive business requirements submitted by users through file upload, model submission, or language interaction, process the business requirements into temporary knowledge objects, call the business scenario intelligent agent workflow, and generate business execution results; when the temporary knowledge object is called by the business scenario intelligent agent workflow and confirmed by the user to have reusable value, upgrade the temporary knowledge object into a formal knowledge object. S6. Write back the business execution results, user feedback, temporary knowledge objects, operation records and new cases to the backend knowledge center to update the knowledge graph library, vector library, structured form library, pattern layer object or knowledge capability node configuration; The backend knowledge center, the business-side intelligent agent workflow, and the frontend application interaction are globally linked through knowledge object identifiers, pattern layer versions, knowledge capability node identifiers, business task identifiers, and operation record identifiers.

2. The knowledge center-driven method for industrial intelligent agents according to claim 1, characterized in that, In S2, the data objects and knowledge objects include: The file object is used to record the file name, file type, file source, uploading user, parsing status, and associated business scenario; The parsed result object is used to record text slices, table fields, drawing areas, model components, rule clauses, and case fields; Vector knowledge objects are used to record vector indexes, semantic representations, and similarity retrieval entry points; Graph knowledge objects are used to record entities, relationships, attributes, constraints, paths between entities, rule nodes, case nodes, rule triggering relationships, case similarity relationships, and subgraph structures; A structured form object is used to record fields, parameters, headers, cells, and structured records, and to establish mapping relationships with entities, attributes, or rule nodes in the graph knowledge object.

3. The knowledge center-driven method for industrial intelligent agents according to claim 2, characterized in that, In S3, the construction of the schema layer object based on the knowledge object specifically includes: The uploaded business data is parsed and segmented to form searchable text, intermediate structure data, and candidate knowledge objects; By combining vector retrieval results, existing references, and user input, entities, relationships, attributes, and rule constraints are constructed in the schema layer canvas; Candidate entities, candidate relationships, candidate attributes, and candidate rules are generated based on associated files; It receives user confirmation, modification, or deletion operations for candidate content and outputs a schema layer object with a version identifier.

4. The knowledge center-driven method for industrial intelligent agents according to claim 3, characterized in that, In S4, the knowledge capability nodes include: The graph retrieval node is used to retrieve entities, relationships, paths, rule nodes, case nodes, subgraphs, and contextual knowledge from the knowledge graph database. Vector retrieval nodes are used to retrieve documents, cases, or drawing fragments that are similar to user requirements, design documents, or model descriptions from the vector library; The rule query node is used to query design criteria, verification rules, or process constraints applicable to the current task stage based on rule nodes, rule relationships, and rule application conditions in the knowledge graph database. Knowledge recommendation nodes are used to generate business suggestions, candidate solutions, or parameter suggestions based on user needs, historical cases, rule nodes, and graph relationships. Knowledge reasoning nodes are used to generate reasoning conclusions based on rule graph relationships, case graph relationships, parameter constraints, structural relationships, or process states. Knowledge update accumulation nodes are used to accumulate newly added knowledge, operation records, business results or case results after user confirmation to the backend knowledge center, and update the corresponding graph nodes, relationships, attributes, subgraphs or vector indexes.

5. A knowledge-center driven system for industrial intelligent agents, applied to the knowledge-center driven method for industrial intelligent agents as described in any one of claims 1 to 4, characterized in that, include: The backend knowledge center module is used to execute S1, S2, and S3, receive multi-source heterogeneous data, and complete data parsing, schema layer construction, knowledge asset organization, and knowledge base updates. The knowledge asset management module is used to execute S3, manage the knowledge graph library, vector library, and structured form library, and manage rule knowledge and case knowledge in the form of graph nodes, relation edges, attributes, and subgraphs; The knowledge capability node management module is used to execute S4, encapsulating graph retrieval, vector retrieval, rule query, knowledge recommendation, knowledge reasoning, and knowledge update accumulation into configurable nodes; The business-side intelligent agent workflow configuration module is used to execute S4 to configure the knowledge capability nodes to one or more task stages in the business scenario intelligent agent workflow. The front-end application interaction module is used to execute S5, receive user-uploaded files, submit models or language interaction input, and display the business execution results; The knowledge update module is used to execute S6 and update the backend knowledge center based on user feedback, running records, temporary knowledge objects, and new cases.

6. A computer device, characterized in that, include: A processor configured to be coupled to memory, read and execute instructions and / or program code in the memory to perform the method as described in any one of claims 1-4.

7. A computer-readable medium, characterized in that, The computer-readable medium stores computer program code that, when executed on a computer, causes the computer to perform the method as described in any one of claims 1-4.

Images