Equipment full-life-cycle management knowledge migration method and system
By using vector similarity matching technology in the equipment management system, the equipment management knowledge is automatically adapted, and the inefficiency problem caused by data differences in cross-factory knowledge migration and expansion is solved, achieving efficient and accurate knowledge sharing.
Patent Information
- Application Number
- CN202411774360.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-05-02
AI Technical Summary
In a multi-factory operation environment, prior art requires additional manpower as data variability in data migration and expansion of equipment management knowledge, resulting in inefficiency.
By obtaining data from the target device, dividing it into text fragments, extracting entity-relational data, building a vector feature model, and using vector similarity matching technology to automatically adapt the data to reduce manual intervention.
It realizes efficient migration and expansion of equipment management knowledge, reduces human resource consumption, and improves the accuracy and efficiency of cross-factory knowledge sharing.
Smart Images
Figure CN119918643A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for equipment full life cycle management knowledge migration. Background Art
[0002] In the manufacturing industry, equipment management is a vital task that covers many complex processes. With the advancement of technology, there are existing cases that show how to use knowledge graphs and big data models to manage equipment inspection, maintenance, repair, outsourcing services, and metrology audits. These cases rely on data foundations to achieve comprehensive management of equipment, and by building knowledge graphs covering equipment, systems, applications, and processes, they improve service quality, optimize customer service processes, and effectively enhance knowledge management capabilities.
[0003] However, in a multi-factory operation environment, these equipment management methods often require additional manpower to build knowledge for new factories and new fields when performing knowledge migration and expansion. For example, when migrating knowledge between different factories, direct migration is not feasible due to the inherent differences in the data of each factory, and the differences must be specially processed. The knowledge graph stores processed structured knowledge, which makes the difference processing based on the original data extremely complicated. For example, when there is equipment management data of a factory (stored in the knowledge graph), if you want to quickly establish an equipment management system for another factory, although you can expand it based on the data of the original factory, you will also encounter the difficulties mentioned in the above scenario during the data expansion process. This shows that in practical applications, although knowledge graphs and big data models provide strong technical support for equipment management, data differences are still a challenge that needs to be overcome in cross-factory knowledge migration and expansion. Summary of the invention
[0004] To this end, the technical problem to be solved by the present invention is to overcome the data differences in the prior art when migrating and expanding knowledge across factories, and to consume additional manpower to build knowledge for new factories and new fields.
[0005] In a first aspect, in order to solve the above technical problems, the present invention provides a method for equipment life cycle management knowledge migration, comprising: Acquire first data from a target device, and segment the first data to obtain a plurality of text segments; Extracting first entity-relationship data according to each of the text segments; Constructing a first vector feature model according to the first entity-relationship data, and acquiring first vector data according to the first vector feature model; The first vector data and the second vector data are matched for similarity using vector similarity; when the similarity is higher than a preset threshold, adaptation data is obtained; when the similarity is lower than the preset threshold, second entity-relationship data is added and adaptation data is obtained; wherein the second vector data and the second entity-relationship data are both obtained from a vector database.
[0006] In one embodiment of the present invention, the method for extracting the first entity-relationship data is: Performing entity recognition on each of the text fragments; determining a relationship between each of said entities; The entity is linked to the vector database, and the entity and the relationship are combined into a triple form.
[0007] In one embodiment of the present invention, the method for determining the relationship between each of the entities is pattern matching or supervised learning or deep learning.
[0008] In one embodiment of the present invention, the method for acquiring the second vector data is: performing entity and relationship recognition on the second data, constructing a second vector feature model, and storing the second vector feature model in the vector database, wherein the second data is acquired from a source device.
[0009] In one embodiment of the present invention, the process of extracting the first entity-relationship data includes using triples of second data as reference, wherein the second data is obtained from a source device.
[0010] In one embodiment of the present invention, the vector database includes a knowledge base.
[0011] In one embodiment of the present invention, the method for constructing the knowledge base is: collecting and integrating information to form a document database; converting the document database into a vector of a preset length; and establishing an index for the vector.
[0012] In a second aspect, in order to solve the above technical problems, the present invention provides a knowledge migration system for equipment life cycle management, comprising: A data slicing module, used for acquiring first data from a target device, and slicing the first data to obtain a plurality of text segments; A data extraction module, configured to extract first entity-relationship data according to each of the text segments; A vector data acquisition module, configured to construct a first vector feature model according to the first entity-relationship data, and acquire first vector data according to the first vector feature model; An adaptation data acquisition module is used to use vector similarity to perform similarity matching between the first vector data and the second vector data; when the similarity is higher than a preset threshold, adaptation data is obtained; when the similarity is lower than the preset threshold, second entity-relationship data is added and adaptation data is obtained again; wherein the second vector data and the second entity-relationship data are both obtained from a vector database.
[0013] In a third aspect, in order to solve the above-mentioned technical problems, the present invention provides an intelligent device, including the above-mentioned equipment full life cycle management knowledge migration system.
[0014] In a fourth aspect, in order to solve the above technical problem, the present invention provides a computer program product, including computer program instructions, which, when executed on a computer, enable any of the above methods to be executed.
[0015] The above technical solution of the present invention has the following beneficial effects compared with the prior art: The method and system for equipment life cycle management knowledge migration described in the present invention realizes the vector representation of data in high-dimensional space by constructing a first vector feature model and obtaining first vector data. This strategy not only greatly facilitates the execution of complex mathematical operations, but also significantly improves the convenience of similarity comparison. With vector similarity as the core, the method can accurately and quickly match the first vector data with the second vector data. When the similarity exceeds the preset threshold, highly similar data can be identified and extracted, thereby significantly improving the matching efficiency. This method allows the rapid application of existing knowledge and speeds up the process of knowledge migration. If the similarity does not reach the preset threshold, the adaptation data will be adjusted and optimized by introducing new second entity-relationship data. This process enhances the flexibility and adaptability of the method and ensures the accuracy and effectiveness of knowledge migration. Therefore, the present invention effectively overcomes the problem of data differences in the process of cross-factory knowledge migration and expansion, and avoids the additional human resource consumption generated by the construction of knowledge systems for new factories and new fields. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein Figure 1 A flowchart of a method for migrating equipment life cycle management knowledge in a preferred embodiment of the present invention; Figure 2 This is an example diagram of the knowledge graph of Factory A in the preferred embodiment of the present invention; Figure 3 This is another example diagram of the knowledge graph of Factory A in the preferred embodiment of the present invention. DETAILED DESCRIPTION
[0017] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention. Embodiment 1
[0018] Reference Figure 1 As shown, an embodiment of the present invention provides a method for migrating equipment lifecycle management knowledge, including: Acquire first data from a target device, and segment the first data to obtain multiple text segments; Extracting first entity-relationship data according to each text segment; Constructing a first vector feature model according to the first entity-relationship data, and acquiring first vector data according to the first vector feature model; Vector similarity is used to perform similarity matching between the first vector data and the second vector data; when the similarity is higher than a preset threshold, adaptation data is obtained; when the similarity is lower than the preset threshold, second entity-relationship data is added and adaptation data is obtained again; wherein the second vector data and the second entity-relationship data are both obtained from a vector database.
[0019] The embodiment of the present invention provides a method for equipment life cycle management knowledge migration, which realizes the vector representation of data in high-dimensional space by constructing a first vector feature model and obtaining first vector data. This representation method is not only convenient for performing complex mathematical operations, but also simplifies the process of similarity comparison. Using vector similarity as a basis, the present embodiment can efficiently match the first vector data with the second vector data. When the similarity exceeds a preset threshold, highly similar data can be identified and extracted, thereby significantly improving the matching efficiency. This method allows the existing knowledge to be applied quickly, which speeds up the process of knowledge migration. If the similarity does not reach the preset threshold, the adaptation data will be adjusted and optimized by introducing new second entity-relationship data. This process enhances the flexibility and adaptability of the method and ensures the accuracy and effectiveness of knowledge migration. Therefore, the embodiment of the present invention effectively overcomes the problem of data differences in the process of cross-factory knowledge migration and expansion, avoids additional human resource consumption, and is used to build a knowledge system for new factories and new fields. Through this method, knowledge between different factories can be smoothly migrated while maintaining the accuracy and applicability of knowledge, providing an innovative solution for equipment management.
[0020] Specifically, for extracting the first entity-relationship data, the specific method is: First, entity recognition is performed on each text segment. The recognition method can be selected from natural language processing technology.
[0021] Secondly, determine the relationship between each entity, and the determination method can be pattern matching, supervised learning, or deep learning. Specifically, pattern matching is used to determine the relationship between each entity, mainly by defining the pattern structure and existing matching knowledge to identify the relationship between entities. This method mainly relies on the limited structural information carried by the pattern itself to mine the pattern matching relationship. For supervised learning to determine the relationship between each entity, it is mainly through the use of labeled data to train the model to identify entities and the relationship between entities. The main steps of this method include: using limited labeled data to train an initial model; using the initial model to predict unlabeled data to obtain prediction results; combining the prediction results with structural information (such as semantic similarity, structural similarity, etc.) to obtain new labeled data; merging the new labeled data with the original labeled data and retraining the model. The advantage of supervised learning is that it can use structural information to improve model performance, but the disadvantage is that it is necessary to calculate the similarity between the prediction results and the structural information, which increases the computational complexity. Deep learning determines the relationship between each entity, mainly including: pipeline learning method (Pipeline), which directly extracts the relationship between entities based on the completion of entity recognition; joint learning method (Joint), based on the end-to-end model of neural network, simultaneously completes the recognition of entities and the extraction of relationships between entities; remote supervised learning method, this method lacks manually annotated data sets, and labels unlabeled data through remote alignment of knowledge bases. The advantage of deep learning is that it can understand the deep semantics and contextual information of text, so as to more accurately identify complex relationships. Each method has its own unique advantages and limitations, and the selection of the appropriate method needs to be determined based on the specific application scenario and available resources.
[0022] Finally, the entities are linked to the vector database, and the entities and relations are combined into the form of triples. In this process, the triples of the second data are used as references, wherein the second data is obtained from the source device.
[0023] Specifically, for the second vector data, the method for obtaining it is as follows: first, the second data is identified for entities and relationships, and a second vector feature model is constructed, wherein the identification method may be a natural language processing (NLP) technology. On the basis of identifying the entities, the interactions and connections between the entities are further analyzed to determine the relationships between them, such as "belonging to", "located in", "made by...", etc. A second vector feature model is constructed based on the identified entities and relationships. This may involve selecting a suitable feature representation method, such as word embeddings, BERT, or other pre-trained language models. Entities and relationships are converted into fixed-length vector representations that can capture the semantic information of entities and relationships in a high-dimensional space. Before storing the vectorized data in the vector database, index optimization is performed to improve retrieval efficiency and performance. The constructed second vector feature model is stored in the vector database to ensure that the organizational structure of the data can support fast retrieval and analysis.
[0024] In the current equipment management model, the common practice is to simply transplant the software deployment and equipment maintenance knowledge of factory A directly to factory B, where factory A is the factory where the source equipment is located and factory B is the factory where the target equipment is located. However, this method ignores the specific needs and environmental differences that may exist in factory B, resulting in the transplanted knowledge not being completely suitable for the actual situation of factory B. In order to solve this problem, an embodiment of the present invention provides a method for efficient and high-quality knowledge migration and expansion. The method can combine the experience of factory A with the environmental characteristics of factory B to provide customized equipment maintenance knowledge for factory B, thereby accelerating the cross-factory knowledge reuse process. Taking the hydraulic pump motor failure of factory B as an example, when encountering the problem that the motor does not run and the hydraulic system cannot output hydraulic oil, we can learn from the experience of factory A. Because factory A may have identified the specific reasons for this phenomenon, including that the motor is not connected to the power supply or the connection is unstable, and there is a problem with the contactor or circuit breaker of the motor. In response to these reasons, factory A provides a series of maintenance methods, such as checking the power connection, confirming the working status of the contactor and circuit breaker, repairing or replacing the faulty parts, and ensuring that all connections are tight and corrosion-free. Through this knowledge transfer method, Factory B was able to quickly apply Factory A’s solution, effectively solve the problem of the hydraulic pump motor not running, and restore the normal operation of the hydraulic system. This method not only improves the efficiency of problem solving, but also ensures the applicability and effectiveness of maintenance measures, thereby improving the overall equipment operation and maintenance quality.
[0025] For the equipment management data of Factory A, the data stored in the knowledge graph is a structured data, and its schema is similar to Figure 2 As shown. Figure 2The Chinese map adopts a hierarchical structure, which clearly depicts the ownership level of equipment management and the logical process of fault handling. In this structure, the group is at the top of the hierarchy, managing multiple factories, and each factory manages specific equipment. When a device fails, a systematic fault handling process is immediately initiated. The process covers everything from recording the fault phenomenon, in-depth analysis of the cause, to formulating solutions, and finally executing maintenance measures.
[0026] In addition to being stored in the knowledge graph, the entity data of the equipment management data of Factory A is also stored in the vector database after extracting feature vectors through the vector model. When data migration or expansion is required in a new factory, first collect the data related to the equipment management of the new factory, such as equipment usage documents, maintenance work orders and other data, and obtain multiple text slices, i.e. text fragments, through the text parsing and segmentation methods commonly used in the large model RAG technology. Use the large language model to extract data for each text slice. When extracting, combine the content of the old factory data to extract the entity-relationship-entity triples, and then perform semantic similarity matching between the extracted entities and the entity data of the old factory equipment management stored in the vector database, so as to disambiguate the entities and relationships, and finally merge them into unified equipment management data for the new and old factories.
[0027] In order to more clearly demonstrate the specific implementation process of the embodiment of the present invention, the following examples will provide a detailed description.
[0028] Example 1: A cable manufacturer in a certain group company is faced with a huge assembly knowledge base, low retrieval efficiency, long employee training cycle, difficulty in knowledge base migration between factories, and inability to form knowledge reuse and knowledge expansion application. Some equipment operation and maintenance data of Factory A is as follows: Figure 3 As shown, at the same time, Factory B has maintenance work order data for the same equipment. Now we will integrate the equipment operation and maintenance management system of Factory A with the maintenance work order data of Factory B to establish an equipment operation and maintenance management data set that can meet the needs of both Factory A and Factory B (this data set is the adaptation data). The specific steps are as follows: Step 1: In addition to being stored in the knowledge graph, the entity and relationship data in the equipment management data of factory A are also stored in the vector database after feature vectors are extracted through a vector model. The vector feature models involved here can be but are not limited to vector feature models such as bge / word2vec / text2vec, and the vector database can be but is not limited to databases that can perform vector storage such as redis / ES.
[0029] Step 2: The data related to the equipment operation and maintenance of factory B (maintenance work orders) are obtained through the commonly used text parsing and segmentation methods in the large model RAG technology, including but not limited to open source text parsing and segmentation methods such as ragflow, to obtain multiple text fragments. The large language model is used to extract data from each text fragment. During the extraction, the entity-relationship-entity triples of factory B are extracted in combination with the equipment operation and maintenance data of factory A, and input into the prompt of the large model. The pseudo code of the extraction process is similar to: “For experienced knowledge graph data extraction personnel, the existing equipment management knowledge graph schema of the xx factory is as follows: {schema} The triple data example of this knowledge graph is: {data example} The known text content to be processed is: {text chunk} Requirements: Extract triple data that conforms to the equipment management knowledge graph schema of the above xx factory from the text content to be processed above, and give it in the following json format: [{"start_v":{"attribute":"attribute content"}, "end_v":{"attribute":"attribute content"}, "relation":{"attribute":"attribute content"}}]. ” Step 3: Perform semantic similarity matching on the entities / relationships extracted by the model and the equipment management entity / relationship data of factory A stored in the vector database. If the similarity is higher than the preset threshold, they are considered to belong to the same entity / relationship and are merged with the original entity / relationship. Otherwise, new entity / relationship data is added, and the relationship of the factory to which each data belongs is added. In this way, the data can be filtered according to the factory to which it belongs to obtain data specific to factory A, data specific to factory B, and data shared by factory A and factory B. Finally, equipment operation and maintenance management data suitable for factory A and factory B is obtained.
[0030] In the application scenario of this example, the device management data processing process was optimized to improve management accuracy and efficiency. First, the feature vector model was used to ensure that the extracted feature vectors can more accurately capture device characteristics and management requirements. Second, the text parsing and data segmentation methods were used to improve the accuracy of extracting useful information from device management documents. In addition, the triple data construction method was used to make the constructed knowledge graph more accurate in representing device management knowledge. These improvements work together to make the system more efficient in processing device management data, provide stronger support for the full life cycle management of equipment, and ensure the quality and reliability of management decisions.
[0031] Furthermore, imagine that a degassing chamber device is associated with two repair orders submitted by a staff member, and each repair order is associated with a different repair order and fault cause. This correlation reflects the complexity of fault handling in equipment management. In order to improve the efficiency of knowledge reuse, these deterministic knowledge are directly migrated to other factories. For knowledge points with general applicability, we allow them to be directly applied in new factories. At the same time, in the daily operation of the factory, new knowledge can be continuously constructed and accumulated, including fault descriptions, fault causes and repair methods, and these knowledge can be stored in the knowledge base in the form of vectors, and the knowledge base is stored in the vector database. Through vector matching technology, deterministic knowledge applicable to the new factory can be identified and extracted, further enriching the knowledge content of equipment management throughout the life cycle. This method not only improves the pertinence and effectiveness of knowledge application, but also enhances the practicality of the knowledge base. In addition, knowledge reuse and derivation can be achieved across industries. For example, the knowledge and experience of the cable industry can be extended to the optical cable industry, and the upstream of the optical fiber industry can also draw on the industrial knowledge of the optical cable industry. This cross-industry knowledge flow was difficult to achieve in the past. But now, with the support of graph databases, vector technology and large models, it is possible to promote cross-border application of knowledge between different industries and realize innovation and optimization of knowledge. In this way, the embodiment of the present invention not only improves the efficiency of knowledge management, but also provides the possibility for knowledge sharing and collaboration between different factories and industries, and promotes the innovation and development of knowledge.
[0032] Exemplarily, in the application scenario of Example 1, the knowledge base is constructed, and the specific steps are as follows: (1) Knowledge base construction: collect and integrate the cable assembly manuals and maintenance manuals of each branch to form a comprehensive document database, including information such as cable models, connector types, tool instructions, safety specifications, and maintenance problem classification. In this example, this information can form about 80,000 pages of documents.
[0033] (2) Knowledge node vectorization: using large models (such as HT-Doc2Vec) to convert documents into vectors of preset length for processing and analysis in high-dimensional vector space. The following is an example of converting nodes into vector representations.
[0034] For example, for a device node, it includes attributes such as device name, model, specification, manufacturer, etc. For example, the vector of a fiber fusion splicer (model: FSM-100, specification: 1550nm, manufacturer: a certain company) is (0.2, 0.5, 0.1, …).
[0035] For example, for a process node, it includes attributes such as process name, process flow, process parameters, etc. For example, the vector of the optical fiber preform preparation process (process flow: raw material selection → batching → melting → drawing → coating → inspection → packaging) is (0.3, 0.4, 0.2, …).
[0036] For example, for a product node, it includes attributes such as product name, model, specification, application scenario, etc. For example, the vector of a multimode optical fiber (model: 62.5 / 125, specification: 850nm, application scenario: data communication) is (0.1, 0.6, 0.3, …).
[0037] For example, for the association relationship, it represents the relationship between equipment, process, and product. The relationship between equipment, process, and product is represented by vectors to support complex queries and analysis. For example, the vector of the optical fiber fusion splicer used in the preparation process of optical fiber preform is (0.8, 0.6, 0.7, …).
[0038] (3) Vector dimension selection: select a vector of appropriate dimension to represent each node document to capture rich semantic information. When selecting the vector dimension, the requirements of the actual application, the complexity of the data, and the limitations of computing resources should be considered comprehensively to select the appropriate vector dimension. In this example, a 300-dimensional vector is selected.
[0039] (4) Index establishment: Use deep learning algorithms (such as Word2Vec and BERT) to establish vector representations for entities and establish full-text search indexes to support fuzzy matching and relevance sorting. For entities such as equipment, process, product, fault description, fault cause, and fault repair method, etc.
[0040] (5) Application expansion allows employees to query the knowledge base by keywords when they encounter problems, and the most relevant knowledge models and solutions will be returned. At the same time, new problems and solutions can be trained through large models to form new knowledge points for retrieval and reference by other branches.
[0041] (6) Performance evaluation: In actual application, the efficiency has been significantly improved. Assembly workers and maintenance personnel have saved an average of 60% of their time in finding solutions, the assembly error rate has been reduced by 20%, and the maintenance efficiency has been improved by about 90%. The parameters 60%, 20%, and 90% are calculated based on this example scenario. The parameters in different application scenarios will be different, and the performance evaluation indicators will also be different. They can be determined based on the actual application situation.
[0042] Through the construction and vectorization of the knowledge base, the intelligence level of cable assembly and maintenance work has been improved, errors have been reduced, and efficiency has been improved. In addition, by sharing and updating the knowledge base, effective knowledge migration and expansion have been achieved. Embodiment 2
[0043] Based on the same inventive concept, this embodiment provides a device life cycle management knowledge migration system, and its principle of solving the problem is similar to the device life cycle management knowledge migration method provided in Example 1, and the repeated parts will not be repeated.
[0044] This embodiment provides a device lifecycle management knowledge migration system, including: A data slicing module, used for acquiring first data from a target device, and slicing the first data to obtain a plurality of text segments; A data extraction module, configured to extract first entity-relationship data according to each text segment; A vector data acquisition module, used to construct a first vector feature model according to the first entity-relationship data, and acquire first vector data according to the first vector feature model; The adaptation data acquisition module is used to use vector similarity to perform similarity matching between the first vector data and the second vector data; when the similarity is higher than a preset threshold, the adaptation data is obtained; when the similarity is lower than the preset threshold, the second entity-relationship data is added and the adaptation data is obtained again; wherein the second vector data and the second entity-relationship data are both obtained from the vector database. Embodiment 3
[0045] This embodiment provides an intelligent device, including the device full life cycle management knowledge migration system provided in the second embodiment. Embodiment 4
[0046] This embodiment provides a computer program product, including computer program instructions. When the computer program instructions are executed on a computer, a device lifecycle management knowledge migration method provided in the first embodiment is executed.
[0047] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0048] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0049] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0050] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0051] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.
Claims
1. A method for equipment life cycle management knowledge migration, characterized in that: include: Acquire first data from a target device, and segment the first data to obtain multiple text segments; Extracting first entity-relationship data according to each of the text segments; Constructing a first vector feature model according to the first entity-relationship data, and acquiring first vector data according to the first vector feature model; The first vector data and the second vector data are matched for similarity using vector similarity; when the similarity is higher than a preset threshold, adaptation data is obtained; when the similarity is lower than the preset threshold, second entity-relationship data is added and adaptation data is obtained; wherein the second vector data and the second entity-relationship data are both obtained from a vector database.
2. A method for equipment life cycle management knowledge migration according to claim 1, characterized in that: The method for extracting the first entity-relationship data is: Performing entity recognition on each of the text fragments; determining a relationship between each of said entities; The entity is linked to the vector database, and the entity and the relationship are combined into a triple form.
3. The method for equipment life cycle management knowledge migration according to claim 2 is characterized in that: The method for determining the relationship between each of the entities is pattern matching, supervised learning, or deep learning.
4. The method for equipment life cycle management knowledge migration according to claim 1, characterized in that: The method for acquiring the second vector data is: performing entity and relationship recognition on the second data, constructing a second vector feature model, and storing the second vector feature model in the vector database, wherein the second data is acquired from a source device.
5. The method for equipment life cycle management knowledge migration according to claim 1, characterized in that: The process of extracting the first entity-relationship data includes using triples of second data as reference, wherein the second data is obtained from a source device.
6. The method for equipment life cycle management knowledge migration according to claim 1, characterized in that: The vector database includes a knowledge base.
7. A method for equipment life cycle management knowledge migration according to claim 6, characterized in that: The method for constructing the knowledge base is as follows: collecting and integrating information to form a document database; converting the document database into a vector of a preset length; and establishing an index for the vector.
8. A knowledge migration system for equipment life cycle management, characterized in that: include: A data slicing module, used for acquiring first data from a target device, and slicing the first data to obtain a plurality of text segments; A data extraction module, configured to extract first entity-relationship data according to each of the text segments; A vector data acquisition module, configured to construct a first vector feature model according to the first entity-relationship data, and acquire first vector data according to the first vector feature model; An adaptation data acquisition module is used to use vector similarity to perform similarity matching between the first vector data and the second vector data; when the similarity is higher than a preset threshold, adaptation data is obtained; when the similarity is lower than the preset threshold, second entity-relationship data is added and adaptation data is obtained again; wherein the second vector data and the second entity-relationship data are both obtained from a vector database.
9. An intelligent device, characterized in that: Including an equipment life cycle management knowledge migration system as described in claim 8.
10. A computer program product, characterized in that The method comprises computer program instructions, which, when executed on a computer, enable the method according to any one of claims 1 to 7 to be executed.