Construction method of insulation board production line fault diagnosis knowledge graph
The construction of the insulating board production line fault diagnosis knowledge graph through knowledge graph technology has solved the problem of low fault diagnosis efficiency in the existing technology, realized complex representation and query of the relationship between fault entities, and improved the accuracy and efficiency of fault diagnosis.
Patent Information
- Application Number
- CN202411045634.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-08
AI Technical Summary
In the prior art, the fault diagnosis of insulation board production lines has problems such as unstructured data cannot be used, fault factor relationships cannot be linked, and lack of knowledge organization and utilization, resulting in low fault diagnosis efficiency.
The knowledge graph technology is used to build the fault diagnosis knowledge graph of the insulation board production line, and the fault diagnosis knowledge graph is realized through pattern construction, knowledge ontology model, named entity relationship recognition model and knowledge storage. The BERT-BiLSTM-CRF model and Neo4j graph database are used to realize the representation and query of fault knowledge.
It improves the accuracy and efficiency of fault diagnosis, can intuitively display the correlation relationship between fault entities, realizes knowledge query from fault causes to repair suggestions, and improves fault diagnosis efficiency.
Smart Images

Figure CN118917397A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for constructing a fault diagnosis knowledge graph, specifically a method for constructing a fault diagnosis knowledge graph for an insulation board production line, and belongs to the technical field of production line fault diagnosis. Background Art
[0002] At present, with the rapid development of economy, the building energy-saving and thermal insulation industry has risen rapidly. The operation status and fault diagnosis of the thermal insulation board production line have a great impact on the production quality and production efficiency of the product.
[0003] At present, in the existing technology, there are some shortcomings in the fault diagnosis of the insulation board production line:
[0004] First, the content of production line failures is mostly unstructured maintenance manuals, semi-structured failure maintenance records and other data, and it is impossible to effectively use historical failure text data;
[0005] Second, the traditional fault diagnosis method cannot link the relationship between various fault factors;
[0006] Third, there is currently a lack of effective organization and utilization of relevant knowledge on fault diagnosis of insulation board production lines.
[0007] Based on this, it is very important to conduct fault diagnosis research. Summary of the invention
[0008] The purpose of the present invention is to provide a method for constructing a knowledge graph for fault diagnosis of an insulation board production line. The construction method applies knowledge graph technology in the field of fault diagnosis, introduces relevant fault knowledge information into the knowledge graph, and describes the fault diagnosis knowledge and its relationships to improve the accuracy of fault diagnosis, so as to solve the problems raised in the above background technology.
[0009] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for constructing a knowledge graph for fault diagnosis of an insulation board production line, the construction method comprising:
[0010] S1. Construction of the knowledge graph for production line fault diagnosis;
[0011] S2. Construction of knowledge ontology model of production line fault diagnosis;
[0012] S3, construction of named entity relationship recognition model;
[0013] S4. Fault knowledge storage.
[0014] As a further technical solution of the present invention: in step S1, the model construction of the production line fault diagnosis knowledge graph adopts a strategy construction model that combines bottom-up and top-down strategies.
[0015] As a further technical solution of the present invention: in step S2, the production line fault diagnosis knowledge ontology model is recorded as: LBPL Fault Diagnosis Ontology; and LBPL Fault Diagnosis Ontology = {Entity, Relation, Attribute};
[0016] in:
[0017] Ⅰ) Entity is the entity set of the production line fault diagnosis knowledge ontology model, and the entity set includes characterization fault, source fault, fault cause, fault location, assembly, maintenance suggestion, maintenance strategy, instrument tool;
[0018] Ⅱ) Relation is the set of relationships between entities and between entities and attributes; denoted as<Entity1,Relation,Entity2> and<Entity,Relation,Attribute> ;
[0019] III) Attribute is a set of attribute concepts, which represents the relevant characteristics of Entity. The attribute concept set includes parameter changes, accessibility, technical attributes and impact scope, and the attribute concept set uses the protégé tool to realize the structuring of the production line fault diagnosis knowledge ontology model.
[0020] As a further technical solution of the present invention: in step S3, the construction of the named entity relationship recognition model includes the following steps:
[0021] S31, data selection, select unstructured text as the data source;
[0022] S32, data preprocessing, marking each text sequence in the fault record, using BIO marking method to mark each character as "BX", "IX" or "O", where B represents the start mark, I represents the middle or end mark, and O represents an irrelevant mark;
[0023] S33, data set division, the text data is divided into training set and test set in a ratio of 7:3 by random division method;
[0024] S34, model training, the training set is input into the BERT-BiLSTM-CRF model for training;
[0025] S35, model verification, inputting a test set of unlabeled data to test the model training results;
[0026] S36. Input the insulation board production line fault text into the model, and output the knowledge triples through the knowledge fusion method.
[0027] As a further technical solution of the present invention: in step S34, the training set is input into the BERT-BiLSTM-CRF model for training, which includes the following steps:
[0028] S341, BERT model converts the input text into an encoding vector;
[0029] S342, the vector output is passed through the BiLSTM model to extract text features and capture the entity relationship before and after the sentence;
[0030] S343, CRF random vector field model corrects the feature matrix transfer probability and outputs the optimal label sequence according to the probability.
[0031] As a further technical solution of the present invention: in step S36, the knowledge fusion method is entity alignment, an entity similarity calculation method based on a combination of cosine similarity and Jaccard correlation coefficient;
[0032] Among them, the calculation method of cosine similarity is:
[0033]
[0034] The calculation method of Jaccard correlation coefficient similarity is:
[0035]
[0036] Where: A and B represent two word vectors, A = {A1, A2, ...An}; B = {B1, B2, ...Bn}.
[0037] As a further technical solution of the present invention: in step S4, the fault knowledge reserve is stored using the Neo4j graph database tool.
[0038] The beneficial effects of the present invention are:
[0039] 1) The present invention uses a knowledge graph to represent the complex relationship between entities, fully utilizes the insulation board production line fault record text data, constructs its fault diagnosis knowledge graph, represents entity knowledge such as characterization of fault, source fault, fault cause, fault location, and maintenance suggestion in the form of a graph, and intuitively displays the association relationship between entities;
[0040] 2) The present invention utilizes the knowledge graph to complete the query of knowledge from fault cause to source fault, source fault to fault phenomenon, fault location to maintenance suggestion, fault phenomenon to fault cause, fault cause to fault location, etc., thereby improving the efficiency of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a schematic diagram of the overall process of the present invention;
[0042] Figure 2 It is a flow chart of fault diagnosis knowledge named entity relationship recognition of the present invention;
[0043] Figure 3 A schematic diagram of a deep learning model for entity extraction according to the present invention;
[0044] Figure 4 This is a visualization result diagram of the fault cause "damage to the agitator blade" in the knowledge graph. DETAILED DESCRIPTION
[0045] In order to facilitate the problem to be solved, an embodiment of the present invention provides a plastic injection molding machine with a demoulding mechanism. The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0046] Examples, such as Figure 1 As shown, the present invention provides a method for constructing a knowledge graph for fault diagnosis of an insulation board production line, and the construction method includes:
[0047] First: Model construction of production line fault diagnosis knowledge graph.
[0048] The model construction of the production line fault diagnosis knowledge graph adopts a strategy construction model that combines bottom-up and top-down approaches.
[0049] Second: Construction of production line fault diagnosis knowledge ontology model.
[0050] In step 2, the production line fault diagnosis knowledge ontology model is recorded as: LBPL Fault DiagnosisOntology;
[0051] And, LBPL Fault Diagnosis Ontology={Entity,Relation,Attribute};
[0052] in:
[0053] Ⅰ) Entity is the entity set of the production line fault diagnosis knowledge ontology model. The entity set defines the entities in the text and divides the entities in the text into: characterization fault, source fault, fault cause, fault location, assembly, maintenance suggestion, maintenance strategy, and instrument tool.
[0054] Specifically: Characterization fault refers to the fault phenomenon that is directly visible during equipment operation; Source fault refers to the fault that is located inside the structure and cannot be directly observed; The cause of the fault is the external factors that are prone to failure; The fault location is the specific part where the fault occurs; The assembly is the fuzzy part where the fault location is located; The maintenance suggestion is the fault handling measure; The maintenance strategy is the practical process of fault maintenance; Instrument tools are the tools used for maintenance.
[0055] Ⅱ) Relation is the set of relationships between entities and between entities and attributes; denoted as<Entity1,Relation,Entity2> and<Entity,Relation,Attribute> .
[0056] Specifically: source fault leads to characteristic fault; fault cause leads to source fault; source fault is located at fault location; assembly is formed at fault location; fault location solution is maintenance suggestion; maintenance measures of assembly are maintenance suggestions; tools of assembly are disassembly tools. The specific relationship is shown in Table 1:
[0057] Table 1 shows the ontology concept relationships.
[0058] Relationship Type Interpretation Leadto (Source fault) causes (Character fault) Cause (Fault reason) causes (source fault) Location (Source fault) is located at (fault location) Component (Assembly) consists of (fault location) Solve (Fault location) Solution is (Maintenance suggestion) Maintenance The maintenance measures for (assembly) are (repair suggestions) Tool The tool for (assembly) is (disassembly tool)
[0059] III) Attribute is a set of attribute concepts, which represents the relevant characteristics of Entity. The attribute concept set includes concepts such as parameter change, accessibility, technical attributes, and impact range. The attribute concept set uses the protégé tool to realize the structuring of the production line fault diagnosis knowledge ontology model.
[0060] Third, the construction of named entity relationship recognition model.
[0061] like Figure 2 As shown, the construction of the named entity relationship recognition model includes the following steps:
[0062] S31, data selection, select the insulation board production line maintenance manual, maintenance records and other unstructured text as the data source;
[0063] S32, data preprocessing, marking each text sequence in the fault record, using BIO marking method to mark each character as "BX", "IX", or "O", where B represents the start mark, I represents the middle or end mark, and O represents an irrelevant mark;
[0064] The entity type tags in the fault record text are shown in Table 2:
[0065] Table 2 Entity class labels
[0066] Entity Type Start tag End Tag Characterizing Failures B-Cfau I-Cfau Source Failure B-Sfau I-Sfau Cause B-Cau I-Cau Fault location B-Loc I-Loc Assembly B-Ass I-Ass Repair suggestions B-Adv I-Adv Maintenance strategy B-Str I-Str Instrument tools B-tol I-tol other O O
[0067] S33, data set division, the text data is divided into training set and test set in a ratio of 7:3 by random division method;
[0068] S34, model training, the training set is input into the BERT-BiLSTM-CRF model for training. The BERT-BiLSTM-CRF model is as follows Figure 3 As shown:
[0069] The training set is input into the BERT-BiLSTM-CRF model for training, which includes the following steps:
[0070] S341, BERT model converts the input text into an encoding vector;
[0071] S342, the vector output is passed through the BiLSTM model to extract text features and capture the entity relationship before and after the sentence;
[0072] S343, CRF random vector field model corrects the feature matrix transfer probability and outputs the optimal label sequence according to the probability;
[0073] S35, model verification, inputting a test set of unlabeled data to test the model training results;
[0074] S36, input the insulation board production line fault text into the model, and output the knowledge triples through the knowledge fusion method;
[0075] The knowledge fusion method is entity alignment, an entity similarity calculation method based on the combination of cosine similarity and Jaccard correlation coefficient;
[0076] Among them, the cosine similarity calculation method is:
[0077]
[0078] The Jaccard correlation coefficient similarity calculation method is:
[0079]
[0080] Where: A and B represent two word vectors, A = {A1, A2, ...An}; B = {B1, B2, ...Bn};
[0081] For the calculation of cosine similarity of word vectors, the closer the value is to 1 (cosθ→1,θ→0), the more similar they are. The Jaccard coefficient describes the similarity of two finite sample sets by defining the ratio of the intersection and union of two sets. The ratio is proportional to the degree of similarity. By calculating the similarity between entities, the candidate entity with the highest similarity is calculated.
[0082] Fourth, fault knowledge storage.
[0083] The fault knowledge reserve uses Neo4j graph database tool for knowledge storage. Figure 4 As shown, it is the visualization result of the atlas fault cause "damage to the stirring tank blade".
[0084] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.
[0085] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.
Claims
1. A method for constructing a knowledge graph for fault diagnosis of an insulation board production line, characterized in that: The construction method comprises: S1. Construction of the knowledge graph for production line fault diagnosis; S2. Construction of knowledge ontology model of production line fault diagnosis; S3, construction of named entity relationship recognition model; S4. Fault knowledge storage.
2. The construction method according to claim 1, characterized in that: In step S1, the model construction of the production line fault diagnosis knowledge graph adopts a strategy construction model that combines bottom-up and top-down approaches.
3. The construction method according to claim 1, characterized in that: In step S2, the production line fault diagnosis knowledge ontology model is recorded as: LBPL Fault Diagnosis Ontology; and LBPL Fault Diagnosis Ontology = {Entity, Relation, Attribute}; in: Ⅰ) Entity is the entity set of the production line fault diagnosis knowledge ontology model, and the entity set includes characterization fault, source fault, fault cause, fault location, assembly, maintenance suggestion, maintenance strategy, instrument tool; Ⅱ) Relation is the set of relationships between entities and between entities and attributes; denoted as<Entity1,Relation,Entity2> and<Entity,Relation,Attribute> ; III) Attribute is a set of attribute concepts, which represents the relevant characteristics of Entity. The attribute concept set includes parameter changes, accessibility, technical attributes and impact scope, and the attribute concept set uses the protégé tool to realize the structuring of the production line fault diagnosis knowledge ontology model.
4. The construction method according to claim 1, characterized in that: In step S3, the construction of the named entity relationship recognition model includes the following steps: S31, data selection, select unstructured text as the data source; S32, data preprocessing, marking each text sequence in the fault record, using BIO marking method to mark each character as "BX", "IX" or "O", where B represents the start mark, I represents the middle or end mark, and O represents an irrelevant mark; S33, data set division, the text data is divided into training set and test set in a ratio of 7:3 by random division method; S34, model training, the training set is input into the BERT-BiLSTM-CRF model for training; S35, model verification, inputting a test set of unlabeled data to test the model training results; S36. Input the insulation board production line fault text into the model, and output the knowledge triples through the knowledge fusion method.
5. The construction method according to claim 4, characterized in that: In step S34, the training set is input into the BERT-BiLSTM-CRF model for training, including the following steps: S341, BERT model converts the input text into an encoding vector; S342, the vector output is passed through the BiLSTM model to extract text features and capture the entity relationship before and after the sentence; S343, CRF random vector field model corrects the feature matrix transfer probability and outputs the optimal label sequence according to the probability.
6. The construction method according to claim 4, characterized in that: In step S36, the knowledge fusion method is entity alignment, an entity similarity calculation method based on a combination of cosine similarity and Jaccard correlation coefficient; Among them, the calculation method of cosine similarity is: The calculation method of Jaccard correlation coefficient similarity is: Where: A and B represent two word vectors, A = {A1, A2, ...An}; B = {B1, B2, ...Bn}.
7. The construction method according to claim 1, characterized in that: In step S4, the fault knowledge reserve is stored using the Neo4j graph database tool.
Citation Information
Patent Citations
Tax domain-oriented knowledge graph construction method
CN111428053A
Knowledge graph construction method for ethylene oxide derivative production process
CN114168745A
Knowledge graph construction method based on reliability of aircraft parts
CN116644192A
Multi-modal fault knowledge graph construction method for belt conveyor
CN117291263A
Cited By
Complex production line reliability analysis method based on combination of fault operation and maintenance knowledge graph and fault tree
CN119903430A
