A radiator process quality knowledge management system based on a knowledge graph
By using a knowledge graph-based radiator process quality knowledge management system, the extraction of process text was optimized, solving the problem of difficulty in sharing and reusing process knowledge, realizing automated annotation and efficient utilization, and improving production efficiency and quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING UNIV
- Filing Date
- 2023-06-25
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies cannot effectively extract process texts, leading to difficulties in sharing and reusing process knowledge, which affects production efficiency and quality.
Design a knowledge graph-based radiator process quality knowledge management system, including process text acquisition, analysis and output modules. Optimize process text extraction through similarity calculation and pattern pool update, and automatically annotate using existing process knowledge.
The process text extraction effect has been optimized, realizing automated annotation and efficient utilization of process knowledge, thereby improving process quality and production efficiency.
Smart Images
Figure CN116737958B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of process knowledge graph technology, and in particular to a heat sink process quality knowledge management system based on knowledge graph. Background Technology
[0002] The sharing and reuse of process knowledge is a crucial aspect of intelligent manufacturing. Process knowledge is not only important for current product design but also valuable for the design of similar products in the future. However, current product process planning heavily relies on expert experience, and it is not arbitrary; it needs to be determined based on specific production conditions such as equipment availability and new product characteristics. Process knowledge is widely distributed, scattered, weakly correlated, and structurally complex and dissimilar. If we can effectively acquire process knowledge and improve its utilization rate, we can significantly reduce process planning costs, increase production efficiency, and improve process quality. This is of great significance for the sharing and reuse of process knowledge and for enhancing process quality.
[0003] Process knowledge differs from general knowledge. Due to its domain-specific vocabulary, diverse knowledge organization characteristics, inconsistent document structure, special semantic combination rules, and strict order requirements, high-quality knowledge cannot be extracted using general knowledge extraction methods. Existing methods are not effective for extracting process text.
[0004] It is evident that designing a knowledge graph-based process quality knowledge management system that can optimize the extraction of process text is a technical problem that urgently needs to be solved. Summary of the Invention
[0005] The technical problem to be solved by this invention is to provide a knowledge graph-based radiator process quality knowledge management system, which is beneficial to optimizing the process text extraction effect.
[0006] To address the aforementioned technical problems, this invention discloses a heat sink process quality knowledge management system based on a knowledge graph. The system includes a process text acquisition module, a first process text analysis module, a second process text analysis module, an output module, and a control module. The process text acquisition module, the first process text analysis module, and the second process text analysis module are each electrically connected to the control module.
[0007] The steps performed by the control module include:
[0008] The control module controls the process text acquisition module to acquire the target process text;
[0009] The control module controls the first analysis module of the process text to determine the target process entity and the type of the target process entity that match the target process text in the pre-built process knowledge base;
[0010] The control module controls the second analysis module of the process text to perform word segmentation on the target process text to obtain the target words corresponding to the target process text.
[0011] The control module controls the second analysis module of the process text to calculate the similarity between the target vocabulary and the vocabulary in the pre-determined pattern pool that matches the target process entity and the type of the target process entity.
[0012] The control module controls the second analysis module of the process text to determine whether the target word is a word outside the pattern pool based on the similarity. If so, the control module controls the output module to add the target word to the pattern pool to form an updated pattern pool, and the control module controls the output module to output a knowledge graph containing the target word.
[0013] In the heat sink process quality knowledge management system disclosed in this invention, the control module controls the second analysis module of the process text to calculate the similarity between the target word and words in a pre-determined pattern pool that match the target process entity and the type of the target process entity. The control module also controls the second analysis module of the process text to determine whether the target word is outside the pattern pool based on the similarity. If so, the control module controls the output module to add the target word to the pattern pool to form an updated pattern pool. The control module also controls the output module to output a knowledge graph containing the target word. This is beneficial for making full use of existing and verified standardized knowledge about heat sink processes to extract process text, for automating annotation, and for optimizing the process text extraction effect.
[0014] As an optional implementation, in this invention, the similarity includes edit distance similarity and cosine similarity.
[0015] As an optional implementation, in this invention, before the control module controls the first analysis module of the process text to determine the target process entity and the target process entity type that match the target process text in the pre-built process knowledge base, the steps performed by the control module further include:
[0016] The control module controls the first analysis module of the process text to generate a pattern pool based on the process knowledge extracted from the target process document and based on regular expressions.
[0017] Furthermore, the process knowledge base includes the pattern pool.
[0018] As an optional implementation, the radiator process quality knowledge management system of the present invention further includes a question content input module, an answer search path generation module, a question content analysis module, and an answer content output module, all electrically connected to the control module.
[0019] The steps performed by the control module also include:
[0020] The control module obtains the text information of the user's question through the question input module;
[0021] The control module controls the question content analysis module to perform feature word detection on the text information and determine the feature words corresponding to the user's question content;
[0022] The control module controls the question content analysis module to filter out target words that match the text information in a pre-determined word set;
[0023] The control module controls the answer search path generation module to generate a single-hop reasoning or multi-hop reasoning answer search path based on the feature words and the target words and meta-paths;
[0024] The control module controls the answer content output module to output the answer content retrieved from a predetermined knowledge graph based on the answer search path.
[0025] As an optional implementation, in this invention, the control module controls the question content analysis module to perform feature word detection on the text information and determine the feature words corresponding to the user's question content, using the Aho-Corasick algorithm. Attached Figure Description
[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0027] Figure 1 This is a schematic diagram of the structure of a heat sink process quality knowledge management system based on a knowledge graph according to an embodiment of the present invention;
[0028] Figure 2 This is a flowchart illustrating the steps executed by the control module in an embodiment of the present invention;
[0029] Figure 3 This is another flowchart illustrating the steps executed by the control module in an embodiment of the present invention;
[0030] Figure 4 This is another flowchart illustrating the steps executed by the control module in an embodiment of the present invention;
[0031] Figure 5 A schematic diagram of single-hop paths and multi-hop paths according to an embodiment of the present invention. Detailed Implementation
[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0033] The terms "first," "second," etc., used in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.
[0034] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0035] This invention discloses a knowledge graph-based knowledge management system for radiator manufacturing processes, applicable to the field of radiator manufacturing. For example... Figure 1 As shown, the radiator process quality knowledge management system includes a process text acquisition module, a process text first analysis module, a process text second analysis module, an output module, and a control module. The process text acquisition module, the process text first analysis module, and the process text second analysis module are electrically connected to the control module.
[0036] like Figure 2 As shown, the steps performed by the control module include:
[0037] S101, the control module controls the process text acquisition module to acquire the target process text. Optionally, the process text acquisition module can acquire the target process text from a process document.
[0038] S102, the control module controls the first analysis module of the process text to determine the target process entities and their types that match the target process text in the pre-built process knowledge base. Step S102 aims to match and extract process entities and their types. For example, entities named "gloves" and "degreasing agent" are directly extracted, and their corresponding entity types (equipment, chemicals) are also extracted and used as new patterns for the next knowledge extraction.
[0039] S103. The control module controls the second analysis module of the process text to perform word segmentation on the target process text to obtain the target words corresponding to the target process text. Optionally, for regular process text, a patterned word segmentation can be performed. Process domain words are mostly expressed in the form of "attribute + headword". LTP word segmentation and part-of-speech tagging are used. If the part of speech of adjacent words after word segmentation is noun, these words are regarded as process domain words and merged.
[0040] S104. The control module controls the second analysis module of the process text to calculate the similarity between the target vocabulary and the vocabulary in the pre-determined pattern pool that matches the target process entity and the target process entity type. Optionally, this similarity may include edit distance similarity and cosine similarity. Further, the calculation of cosine similarity cos(s,t) can be found in the following formula:
[0041]
[0042] In the formula, s is the vectorized representation of the string corresponding to the words that match the target process entity and the target process entity type in the pre-determined pattern pool, and t is the vectorized representation of the string of the target word.
[0043] Furthermore, the calculation of edit distance similarity ED(s,t) can be found in the following formula:
[0044]
[0045] In the formula, dis(s,t) is the edit distance between the string corresponding to the word that matches the target process entity and the target process entity type in the pre-determined pattern pool and the string of the target word, and Len(s) and Len(t) are the lengths between the string corresponding to the word that matches the target process entity and the target process entity type in the pre-determined pattern pool and the string of the target word.
[0046] S105. The control module controls the second analysis module of the process text to determine whether the target word is outside the pattern pool based on similarity. If yes, proceed to step S106; optionally, if no, return to step S101. Optionally, in step S105, the similarity can be greater than or equal to a preset threshold to determine whether the target word is outside the pattern pool. Specifically, the edit distance threshold is set to w1, and the cosine similarity threshold is set to w2. If the edit distance and cosine similarity are higher than w1 and w2 respectively, it is determined whether the target word is outside the pattern pool and whether its concept is the same as the concept of the entity it is compared with.
[0047] S106. The control module controls the output module to add the target vocabulary to the pattern pool, forming an updated pattern pool, and the control module controls the output module to output a knowledge graph containing the target vocabulary.
[0048] As can be seen, in the radiator process quality knowledge management system based on knowledge graph disclosed in this invention, the control module controls the second analysis module of the process text to calculate the similarity between the target word and words in the pre-determined pattern pool that match the target process entity and the target process entity type. The control module also controls the second analysis module of the process text to determine whether the target word is outside the pattern pool based on the similarity. If so, the control module controls the output module to add the target word to the pattern pool to form an updated pattern pool. The control module also controls the output module to output a knowledge graph containing the target word. This is beneficial for making full use of existing and verified standardized knowledge about radiator processes to extract process text, for automating annotation, and for optimizing the process text extraction effect.
[0049] To further optimize the extraction of process text, preprocessing can be performed on the process text before the control module controls the first analysis module to determine the target process entity and its type that match the target process text in the pre-built process knowledge base. Specifically, this preprocessing occurs before step S102, such as... Figure 3 As shown, the steps performed by the control module also include:
[0050] S1011, The control module controls the first analysis module of the process text, which generates a pattern pool based on the process knowledge extracted from the target process document using regular expressions.
[0051] In addition, the process knowledge base mentioned in step S102 includes a pattern pool.
[0052] Optionally, the content of the process text can be extracted using regular expressions from template-based process texts and stored in a knowledge graph. Entities and their concepts, together, serve as seed patterns stored in a pattern pool to facilitate subsequent knowledge extraction from unstructured texts. Further, optionally, unstructured texts can use regular expressions to identify keywords as the text's topic. If the topic is sequential knowledge, fine-grained extraction and automatic generation of sequential relationships are performed, with the order identified by numbers.
[0053] To more specifically describe the implementation process of the second analysis module of the control process text of the control module in step S104, which calculates the similarity between the target words and the words in the pre-determined pattern pool that match the target process entity and the target process entity type, and the second analysis module of the control process text of the control module in step S105, which determines whether the target words are words outside the pattern pool based on the similarity, the following will describe the relevant details using a specific application scenario as an example.
[0054] After word segmentation is completed, the following two situations can be identified.
[0055] (1) For target words that contain only core words, directly calculate the edit distance similarity and cosine similarity scores between them and existing entities, and compare the two results with the threshold. If the word is only part of an entity, such as "rubber frame", but it is only a subset of "yellow rubber frame", then "rubber frame" is deleted because it is not an accurate representation of the term in the manufacturing process domain.
[0056] (2) For target words that contain both core words and attributes, the similarity calculation is divided into two parts: attribute similarity calculation and core word similarity calculation. Attribute similarity score is obtained by calculating the similarity of each attribute contained in the word, calculating the edit distance similarity and cosine similarity of each attribute separately, and then averaging them separately. Core word similarity is calculated by calculating the edit distance similarity and cosine similarity of words other than attributes. In the field of manufacturing processes, we usually use the last meaningful word in a phrase as the core word. After obtaining the edit distance similarity and cosine similarity of attributes and core words respectively, a weight of 0.5 is assigned to attributes and core words respectively to obtain the edit distance and cosine similarity of the entire word.
[0057] Based on multiple sets of experimental data, w1 and w2 were set to 0.5. For example, "neatly place semi-finished products in blue rubber frames and neatly place defective products in yellow rubber frames" is an unstructured process text. There is a pattern <equipment: blue rubber frame> in the pattern pool. The entity "blue rubber frame" is extracted using entity-level masking, and the word "yellow rubber frame" is obtained through word segmentation in character-level masking, where "yellow" is an attribute and "rubber frame" is the core word. Using this method, the edit distance similarity and cosine similarity of the entire word can be calculated to be 0.583 and 0.530, respectively, both higher than w1 and w2. Therefore, "yellow rubber frame" is added to the pattern pool and the process knowledge graph.
[0058] To further enrich the functionality of the knowledge graph-based radiator process quality knowledge management system, and to fully utilize the radiator process knowledge base or knowledge graph formed through the optimized knowledge extraction process described above, this invention can also provide corresponding answers to user questions. Specifically, such as... Figure 1 As shown, the radiator process quality knowledge management system also includes a question content input module, an answer search path generation module, a question content analysis module, and an answer content output module, all of which are electrically connected to the control module.
[0059] like Figure 4 As shown, the steps performed by the control module also include:
[0060] S201. The control module obtains the text information of the user's question through the question input module.
[0061] S202, the control module controls the question content analysis module to perform feature word detection on the text information and determine the feature words corresponding to the user's question content. The Aho-Corasick algorithm can be used to identify feature words in the user's question. The Aho-Corasick algorithm detects multiple patterns in the text. It only needs to process the input question once to find all the keywords that appear, even if they overlap. In this way, all features appearing in the question will be detected.
[0062] S203, the control module controls the question content analysis module to filter target words that match the text information in a pre-determined word set. In this step, the presence of words from the word set in the user's question can be determined through the text description, and these are considered target words. Furthermore, to address the issue of failing to identify potential target words in the question due to overly strict word matching, a strategy of expanding the target word set through synonyms and words the user might ask can be adopted to relax the word matching conditions. Specifically, detecting feature words and target words in the question can determine the user's intent. Further, optionally, after detecting feature words and target words, the following steps can be taken:<features-targets>The form represents the intent of the problem and is used for path matching in the design.
[0063] S204. The control module controls the answer search path generation module to generate single-hop or multi-hop reasoning answer search paths based on feature words and target words using meta-paths. Here, a meta-path is a series of relationships connecting two objects. Optionally, on a meta-path, single-hop reasoning only requires one hop from the starting node to the target node, while multi-hop reasoning requires traversing multiple nodes from the starting node to reach the target node after multiple hops. Based on the constructed process ontology, the concepts and relationships in the ontology are used as nodes and edges of the meta-path. This method extends the meta-path-based search path. According to complexity, the search path is divided into single-hop and multi-hop paths, and the path is designed to be represented in the form of <feature-target>. Figure 5 As shown, the forms of single-hop paths and multi-hop paths are <workpiece-process> and <workpiece-process-notes>, respectively.
[0064] S205, The control module controls the answer content output module to output the answer content retrieved from a predetermined knowledge graph based on the answer search path.
[0065] Finally, it should be noted that the knowledge graph-based radiator process quality knowledge management system disclosed in this embodiment of the invention is merely a preferred embodiment of the invention and is only used to illustrate the technical solution of the invention, not to limit it. Although the invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the invention.
Claims
1. A knowledge graph-based knowledge management system for radiator manufacturing process quality, characterized in that, The radiator process quality knowledge management system includes a process text acquisition module, a process text first analysis module, a process text second analysis module, an output module, and a control module. The process text acquisition module, the process text first analysis module, and the process text second analysis module are each electrically connected to the control module. The steps performed by the control module include: The control module controls the process text acquisition module to acquire the target process text; The control module controls the first analysis module of the process text to determine the target process entity and the type of the target process entity that match the target process text in the pre-built process knowledge base; The control module controls the second analysis module of the process text to perform word segmentation on the target process text to obtain the target words corresponding to the target process text. The control module controls the second analysis module of the process text to calculate the similarity between the target vocabulary and the vocabulary in the pre-determined pattern pool that matches the target process entity and the type of the target process entity. The control module controls the second analysis module of the process text to determine whether the target word is a word outside the pattern pool based on the similarity. If so, the control module controls the output module to add the target word to the pattern pool to form an updated pattern pool, and the control module controls the output module to output a knowledge graph containing the target word. Before the control module controls the first process text analysis module to determine the target process entity and target process entity type that match the target process text in the pre-built process knowledge base, the steps performed by the control module further include: the control module controls the first process text analysis module to generate a pattern pool based on the process knowledge extracted from the target process document using regular expressions; entities are extracted from the template-based process text using regular expressions and stored in the knowledge graph, and the entities and their concepts are stored as seed patterns in the pattern pool; the process knowledge base includes the pattern pool.
2. The knowledge graph-based radiator process quality knowledge management system according to claim 1, characterized in that, The similarity includes edit distance similarity and cosine similarity.
3. The knowledge graph-based radiator process quality knowledge management system according to claim 2, characterized in that, The radiator process quality knowledge management system also includes a question content input module, an answer search path generation module, a question content analysis module, and an answer content output module, all electrically connected to the control module. The steps performed by the control module also include: The control module obtains the text information of the user's question through the question input module; The control module controls the question content analysis module to perform feature word detection on the text information and determine the feature words corresponding to the user's question content; The control module controls the question content analysis module to filter out target words that match the text information in a pre-determined word set; The control module controls the answer search path generation module to generate a single-hop reasoning or multi-hop reasoning answer search path based on the feature words and the target words and meta-paths; The control module controls the answer content output module to output the answer content retrieved from a predetermined knowledge graph based on the answer search path.
4. The knowledge graph-based radiator process quality knowledge management system according to claim 3, characterized in that, The control module controls the question content analysis module to perform feature word detection on the text information and determine the feature words corresponding to the user's question content, using the Aho-Corasick algorithm.