Industrial robot welding process-oriented knowledge graph construction method and related system thereof

Abstract

The invention provides an industrial robot welding process-oriented knowledge graph construction method and a related system thereof, and belongs to the technical field of knowledge graphs. Specifically, industrial robot welding process data are collected, and an industrial robot welding process data database is established; constructing a welding knowledge graph mode layer from top to bottom based on a welding process data database; constructing a knowledge graph data layer from bottom to top based on a welding knowledge graph mode layer, and forming an entity-relationship-entity triple data form by adopting a knowledge extraction algorithm; aiming at an entity-relation-entity triple data form, performing data alignment by adopting a similarity algorithm, and merging equivalent entities and relations; and storing all the triple data, and importing the triple data into a Neo4j graph database to realize visualization of the knowledge graph. Reliability and efficiency of the industrial robot welding process can be improved, and agility and intelligence of the industrial robot are promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of knowledge graphs, and specifically relates to a method for constructing a knowledge graph for industrial robot welding processes and a related system thereof. Background Art

[0002] Industrial robot welding is a key technology in modern manufacturing. Due to its high efficiency, high precision and high degree of automation, it is widely used in many industries such as automobile manufacturing, aerospace, shipbuilding, and machinery manufacturing. As the global manufacturing industry continues to increase its requirements for product quality, production efficiency, and cost control, welding robots have gradually replaced traditional manual welding in industrial production and become an important part of intelligent manufacturing.

[0003] Knowledge graph is an emerging artificial intelligence technology. By building a structured semantic network, it can effectively integrate and manage massive amounts of heterogeneous data, helping to achieve knowledge representation, organization and reasoning. In recent years, the application of knowledge graphs in the field of mechanical manufacturing has become increasingly widespread, becoming an important tool for improving the level of intelligent manufacturing and decision support. The mechanical manufacturing process is complex and involves a large number of different types of data, including design drawings, processing technology, equipment parameters, quality standards, fault diagnosis information, etc. These data are usually scattered in different systems and are difficult to effectively associate and use, resulting in serious information islands. Knowledge graph technology semantically associates data from different sources to build a comprehensive and dynamic knowledge network in the processing field, realizing information integration and intelligent processing.

[0004] In order to promote the development of industrial robots towards agility and intelligence, knowledge graph technology is applied to the field of industrial robots. Constructing a knowledge graph for industrial robot welding technology can systematize and structure the data and experience knowledge involved in the welding process, optimize the welding process, improve the reliability and processing efficiency of welding, and promote the refinement and intelligence of welding. Summary of the invention

[0005] The purpose of the present invention is to overcome the problems of complex current industrial robot welding process data, many influencing factors, low digitization rate of knowledge and experience, serious information fragmentation, and to provide a method for constructing an industrial robot welding process knowledge graph and its related system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for constructing a welding process knowledge graph for an industrial robot, comprising the following steps: Collect industrial robot welding process data and establish an industrial robot welding process data database; Based on the welding process data database, the welding knowledge graph model layer is constructed from top to bottom; Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the knowledge extraction algorithm is used to form the entity-relationship-entity triple data form; For the entity-relationship-entity triple data form, a similarity algorithm is used to align the data and merge equivalent entities and relationships; All triple data are stored and imported into the Neo4j graph database to realize the visualization of the knowledge graph.

[0007] The specific method of constructing the welding knowledge graph model layer from top to bottom is as follows: Analyze the welding process types and specific welding processes of industrial robots, and construct corresponding entities according to different welding process types as welding process classes; For each different welding process category, divide it into subcategories and set the relevant attributes of the subcategories; Define the types of different welding process classes and the hierarchical relationship of their subclasses to complete the top-down model layer construction.

[0008] The industrial robot welding process data database includes structured data and unstructured data, wherein the structured data is welding objects, specific welding processes and welding process parameters, and the unstructured data is text sentences with unfixed format and length.

[0009] Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the specific method of using the knowledge extraction algorithm to form the entity-relationship-entity triple data form is as follows: Post-process the structured data to form a triple data form of entity-relationship-entity; Annotate unstructured data; The labeled data is divided into a training set, a test set, and a validation set, the Transformer model is trained based on the training set, the model parameters in the training process are adjusted based on the validation set, and the Transformer model is tested based on the test set to obtain a trained Transformer model; Load the trained Transformer model to extract knowledge from unlabeled data, post-process the entity label sequence, and obtain data in the form of entity-relationship-entity triples.

[0010] The specific method for labeling unstructured data is as follows: Use OCR algorithm to identify text information of unstructured data and convert it into txt files; Perform data cleaning on the text to remove noise in the text; According to the welding process entities constructed in the pattern layer, the text is annotated using the BIOES annotation strategy to distinguish different welding process entities; Define the relationship between entities in the welding process and use the same annotation strategy for annotation.

[0011] The FLAT algorithm is used to train the Transformer model. The specific method is as follows: Perform word embedding and position encoding on the input text; Train through the self-attention mechanism layer to calculate the correlation between characters and words; Perform residual connection and normalization operations on the training results to stabilize the training results; Perform nonlinear transformation on the training results to obtain transformation results; Apply addition and standardization operations to the transformation results to stabilize the output of the model; Convert the transformation result into the final output result and output the entity label sequence.

[0012] The similarity algorithm based on edit distance is used for data alignment. A threshold is set. When the calculation result is greater than the threshold, the two entities are considered to be overlapping and then merged. The specific calculation formula is as follows:

[0013] in, Indicates the first string i characters with the first character of the second string j The edit distance between characters; Represents a replacement operation. Indicates a delete operation. Represents an insert operation.

[0014] In a second aspect, the present invention provides a system for constructing a welding process knowledge graph for an industrial robot, comprising: Data collection module, used to collect industrial robot welding process data and establish an industrial robot welding process data database; The model layer construction module is used to construct the welding knowledge graph model layer from top to bottom based on the welding process data database; The data layer construction module is used to construct the knowledge graph data layer from the bottom up based on the welding knowledge graph model layer, and adopt the knowledge extraction algorithm to form the entity-relationship-entity triple data form; The data alignment module is used to align the entity-relationship-entity triple data using a similarity algorithm and merge equivalent entities and relationships. The storage and visualization module is used to store all triple data and import them into the Neo4j graph database to realize the visualization of the knowledge graph.

[0015] In a third aspect, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of a method for constructing a welding process knowledge graph for an industrial robot.

[0016] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for constructing a welding process knowledge graph for an industrial robot.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for constructing a welding process knowledge graph for an industrial robot and a related system thereof. Specifically, the welding process data of the industrial robot is collected, and a welding process data database of the industrial robot is established; based on the welding process data database, a welding knowledge graph pattern layer is constructed from top to bottom; based on the welding knowledge graph pattern layer, a knowledge graph data layer is constructed from bottom to top, and a knowledge extraction algorithm is used to form a triple data form of entity-relationship-entity; for the triple data form of entity-relationship-entity, a similarity algorithm is used to align data, and equivalent entities and relationships are merged; all triple data are stored and imported into a Neo4j graph database to realize the visualization of the knowledge graph. By forming the fragmented process data into a triple data form through a knowledge extraction algorithm, the triple data is imported into a Neo4j graph database to realize the visualization of the knowledge graph, and the robot automatically adjusts the welding parameters, such as current, voltage, welding speed and welding angle, according to the best practices in the knowledge graph, so as to maximize the stability and accuracy of the welding process. This not only reduces the error caused by manual intervention, but also effectively reduces the welding defect rate.

[0018] Furthermore, the industrial robot welding process knowledge map brings together key technical data and practical experience in the welding process, and can be continuously updated and expanded. This knowledge base can not only provide a complete technology accumulation platform for enterprises, but also promote technology sharing and communication within enterprises and between industries. Through standardized and systematic knowledge management, enterprises can quickly transfer complex welding process experience and reduce training costs. At the same time, it lays a solid foundation for the optimization and innovation of welding technology, and effectively promotes the overall technological progress of the welding industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the present invention; Figure 2A schematic diagram of the construction of the welding process ontology model of the industrial robot of the present invention; Figure 3 It is a schematic diagram of the structure of the knowledge extraction algorithm FLAT of the present invention; Figure 4 Generate a triple data flow diagram for the present invention; Figure 5 This is the partial result of the visualization of the welding process knowledge graph of the process robot of the present invention.

[0020] Figure 6 It is a system diagram of the present invention; Figure 7 This is a system diagram of Example 3 of the present invention. DETAILED DESCRIPTION

[0021] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0022] Embodiment 1: A method for constructing a knowledge graph for industrial robot welding process includes the following steps: S1: Collect industrial robot welding process data and establish an industrial robot welding process data database; S2: Based on the welding process data database, the welding knowledge graph model layer is constructed from top to bottom; S3: Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the knowledge extraction algorithm is used to form the entity-relationship-entity triple data form; S4: For the entity-relationship-entity triple data form, a similarity algorithm is used to align the data and merge equivalent entities and relations; S5: All triple data are stored and imported into the Neo4j graph database to visualize the knowledge graph.

[0023] Specifically, in S1, industrial robot welding process data is collected, mainly including industrial robot welding process manuals and industrial robot welding process parameters formed by enterprise production practices. The industrial robot welding process manual involves welding types, welding process and other contents. For other welding knowledge, such as the material of the welding object, you can refer to the welding process manual. In addition to the above information, you can also collect papers and monographs on industrial robot welding as knowledge supplements, and collect expert experience as guidance. Based on the above content, a data database of the knowledge graph is established to prepare for the construction of the knowledge graph model layer and data layer.

[0024] Specifically, in S2, the specific method of constructing the welding knowledge graph model layer from top to bottom is as follows: S21: Analyze the welding types and specific welding processes of industrial robots, and build classes according to different welding process types. Each class represents a category of welding process or a specific welding concept. Then, based on the class, further subclasses are divided, and relevant properties are set for each subclass, including object properties and data properties. These properties can describe the specific parameters, equipment characteristics and operation details of the welding process in detail, thereby forming a complete welding process knowledge model.

[0025] S22: Define the ontology types and hierarchical relationships of industrial robot welding. In the ontology construction of industrial robot welding process, the main contents include: welding type, welding object, welding scene and welding process parameters. For welding type, there are currently 8 main types suitable for industrial robot welding, namely: arc welding, plasma welding, spot welding, laser welding, resistance welding, friction welding, ultrasonic welding and laser welding process. When constructing the ontology, first establish the top-level ontology of "welding process type", and then further construct the subclasses of these 8 welding types. For example, for the subclass of arc welding, which is widely used in the industry, its sub-methods include arc welding, submerged arc automatic welding, argon arc welding and CO2 gas shielded welding. Taking arc welding as an example, as a subclass of arc welding process, it involves many aspects of process knowledge, including the form of welding joints (such as butt joints, corner joints, etc.), welding process (such as starting, ending, etc.), and welding methods for different welding positions (such as flat welding, vertical welding, overhead welding, etc.). Define the knowledge of each part in detail in the ontology, and ensure their mutual association and logical consistency.

[0026] S23: Ontology is also constructed for other contents involved in the welding process of industrial robots. With the help of the tool protégé, the top-down model layer construction is completed to realize the definition of ontology type, object attributes, data attributes and the visualization of ontology.

[0027] Specifically, in S3, the welding process data resource library contains a large amount of process knowledge. The data layer is constructed by extracting structured process knowledge (such as standard process flow, process parameter table, etc.) and unstructured process knowledge (such as welding experts' experience records, text descriptions in documents, reports), and converting them into triples in the knowledge graph.

[0028] The specific method of building the knowledge graph data layer from the bottom up is as follows: S31: Structured data can be directly extracted from databases, spreadsheets or other data sources through Python tools or other data extraction tools (such as Pandas, CSV library, etc.) due to its standardized and clear format. These structured data usually exist in the form of tables or database records, with clear fields and relationships. In the process of building the knowledge graph, Python programs are used to process these data and finally output them in the form of triples that meet the requirements of the knowledge graph. Take a specific example: In the welding rod arc welding process table, the joint groove form is recorded as "butt joint I shape", and the applicable scenario is "plate thickness less than 4mm, no groove". Through Python's data extraction operation, these structured information can be identified and converted into a triple form of <joint groove form: butt joint I shape-applicable to-scenario: plate thickness less than 4mm, no groove>.

[0029] S32: Unstructured data mainly comes from welding process documents, reports, expert experience records, etc. These data are usually descriptive statements of physical phenomena in the welding process, or empirical conclusions formed through long-term practice. Unstructured data usually exists in the form of text and lacks a fixed structure or format. This type of data cannot be directly used for knowledge extraction. It needs to be converted into triples through a series of operations such as text extraction, data cleaning, entity annotation, model training, and post-processing.

[0030] Specifically, text extraction uses the OCR algorithm in the python tool to identify text information in the process document PDF, and then batch converts it into txt files. The extracted text is cleaned to remove noise in the text, such as irrelevant text, typos, special symbols, etc., and standardize the text format. Before entity annotation, the entity is defined first, and the BIOES annotation strategy is adopted according to the ontology model constructed in the pattern layer. B represents the first element of the entity, I represents the element in the middle of the entity, E represents the element at the end of the entity, O represents the non-entity type, and S represents the entity length of 1. For example, in the entity "stainless steel", "not" is the beginning element of the entity and is marked as B, "rust" is the element in the middle of the entity and is marked as I, and "steel" is marked as E. For single element entities, such as "aluminum", it is marked as S. Since the welding process pattern layer contains multiple ontologies, the derived entity types are also diverse. In order to distinguish entities belonging to different categories, suffixes are added to the labels during annotation. For example, the "stainless steel" entity mentioned above belongs to the material category in welding, which is marked as B-MatR, I-MatR, E-MatR, where MatR represents the material. Steel, copper, aluminum alloy, etc. all belong to this type of entity. The definitions of other types of entities are shown in Table 1.

[0031] In addition to defining the entities in the welding process, it is also necessary to define the relationship between entities in the welding process, adopting the same annotation strategy. The specific relationship definition is shown in Table 2.

[0032] S33: Use the annotated text corpus as the data set for model training, and divide the data set into training set, validation set and test set. The model uses the Chinese named entity recognition algorithm FLAT under the Transformer framework. Its algorithm structure is as follows: Figure 3 The specific working process is as follows: 1) The input text is embedded in the embedding layer, and the input data is encoded in position. FLAT uses relative position encoding to encode the position of each word in the input token in the sentence and the position of the first and last words in the sentence, so as to understand the complex positional relationship between words; 2) Training is performed through the self-attention layer to calculate the correlation between characters and words, understand the semantics, and obtain training results; 3) Perform residual connection and normalization operations on the training results by adding the Add&Norm layer to stabilize the training results; 4) Perform nonlinear transformation on the training results through the feedforward neural network (FFN) layer to obtain the transformation results; the feedforward neural network layer FFN includes two linear layers and one nonlinear activation layer; 5) Apply addition and normalization operations to the transformed results again through the Add&Norm layer to further stabilize the output of the model; 6) The output of the feedforward neural network is converted into the final output result through the linear and CRF (Linear & CRF) layers, and the entity label sequence is output.

[0033] In this model training process, the learning rate is set to 0.01, the epoch is set to 50, the batch size is set to 16, and the number of intermediate nodes of the feedforward neural network is set to 480. The trained model is tested using the test set, and the accuracy is 93%. The model parameters are saved for welding process knowledge extraction.

[0034] S34: Figure 4As shown in the figure, the trained model is loaded to extract knowledge from the unlabeled text corpus. Taking "Brazing is widely used in the connection of stainless steel" as an example, the text input model has not been labeled before, belongs to unstructured data, and cannot directly form triples. Through the FLAT algorithm, each word in the text is predicted and its label is output. According to the results, in this sentence, "brazing" and "stainless steel" are entities, belonging to the welding process entity and welding material entity respectively, and "applied to" belongs to a relationship. On this basis, the output results are post-processed using the python program to obtain the data in the form of triples "brazing-applied to-stainless steel".

[0035] Specifically, in S4, for the entity-relationship-entity triple data form, a similarity algorithm is used to align data and merge equivalent entities and relationships. The data for knowledge extraction comes from different process files, and the standardization of entity names cannot be guaranteed. There may be different names for the same entity. In order to eliminate overlapping features, achieve entity alignment, and ensure the standardization and uniformity of entity names, a similarity calculation method based on edit distance is used.

[0036] 1) Levenshtein distance refers to the minimum number of operations required to convert one string into another. Operations include addition, deletion, and modification. Generally, the smaller the Levenshtein distance, the higher the similarity between the two strings. The specific calculation formula is as follows:

[0037] in, Indicates the first string i characters with the first character of the second string j The edit distance between characters. Represents a replacement operation. Indicates a delete operation. Represents an insert operation.

[0038] 2) The similarity between entities can be obtained by calculating the edit distance. A threshold is set. When the calculation result is greater than the threshold, it can be determined that the two entities are overlapping and then merged. See Table 3 for specific examples.

[0039] Specifically, in S5, the relational database MySQL is used to store all triple data that have undergone the knowledge fusion process. In order to further realize the visualization of the knowledge graph, the graph database Neo4j is used to store and visualize all triples. Neo4j is a database based on a graph structure that can more intuitively represent the connection between entities and relationships in the knowledge graph. By exporting the triple data stored in MySQL and then importing it into the Neo4j database, the complex structure of the welding process knowledge can be displayed in a graphical way.

[0040] like Figure 5 As shown in Figure 1, in Neo4j, the entities (subject and object) in each triple correspond to the nodes in the graph, and the relationships (predicates) in the triples correspond to the edges between the nodes. Through the graphical interface of the graph database, various welding types, objects, parameters and their relationships in the welding process are clearly presented, allowing users to quickly obtain the associated information and knowledge structure of the welding process from the graph.

[0041] Embodiment 2: like Figure 6 As shown, a system for constructing a knowledge graph for industrial robot welding process includes: Data collection module, used to collect industrial robot welding process data and establish an industrial robot welding process data database; The model layer construction module is used to construct the welding knowledge graph model layer from top to bottom based on the welding process data database; The data layer construction module is used to construct the knowledge graph data layer from the bottom up based on the welding knowledge graph model layer, and adopt the knowledge extraction algorithm to form the entity-relationship-entity triple data form; The data alignment module is used to align the entity-relationship-entity triple data using a similarity algorithm and merge equivalent entities and relationships. The storage and visualization module is used to store all triple data and import them into the Neo4j graph database to realize the visualization of the knowledge graph.

[0042] Embodiment 3: like Figure 7 As shown, the present invention also provides an electronic device 100 for a method for constructing a welding process knowledge graph for an industrial robot; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0043] The memory 101 can be used to store the computer program 103, and the processor 102 implements the steps of the method for constructing a welding process knowledge graph for industrial robots described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0044] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0045] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for constructing a welding process knowledge graph for an industrial robot, and the processor 102 can execute the plurality of instructions to implement: Collect industrial robot welding process data and establish an industrial robot welding process data database; Based on the welding process data database, the welding knowledge graph model layer is constructed from top to bottom; Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the knowledge extraction algorithm is used to form the entity-relationship-entity triple data form; For the entity-relationship-entity triple data form, a similarity algorithm is used to align the data and merge equivalent entities and relationships; All triple data are stored and imported into the Neo4j graph database to realize the visualization of the knowledge graph.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a knowledge graph for industrial robot welding process, characterized in that: The steps include: Collect industrial robot welding process data and establish an industrial robot welding process data database; Based on the welding process data database, the welding knowledge graph model layer is constructed from top to bottom; Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the knowledge extraction algorithm is used to form the entity-relationship-entity triple data form; For the entity-relationship-entity triple data form, a similarity algorithm is used to align the data and merge equivalent entities and relationships; All triple data are stored and imported into the Neo4j graph database to realize the visualization of the knowledge graph.

2. A method for constructing a knowledge graph for industrial robot welding process according to claim 1, characterized in that: The specific method of constructing the welding knowledge graph model layer from top to bottom is as follows: Analyze the welding process types and specific welding processes of industrial robots, and construct corresponding entities according to different welding process types as welding process classes; For each different welding process category, divide it into subcategories and set the relevant attributes of the subcategories; Define the types of different welding process classes and the hierarchical relationship of their subclasses to complete the top-down model layer construction.

3. The method for constructing a knowledge graph for industrial robot welding process according to claim 1, characterized in that: The industrial robot welding process data database includes structured data and unstructured data, wherein the structured data is welding objects, specific welding processes and welding process parameters, and the unstructured data is text sentences with unfixed format and length.

4. A method for constructing a knowledge graph for industrial robot welding process according to claim 3, characterized in that: Based on the welding knowledge graph model layer, the knowledge graph data layer is constructed from the bottom up, and the specific method of using the knowledge extraction algorithm to form the entity-relationship-entity triple data form is as follows: Post-process the structured data to form a triple data form of entity-relationship-entity; Annotate unstructured data; The labeled data is divided into a training set, a test set, and a validation set, the Transformer model is trained based on the training set, the model parameters in the training process are adjusted based on the validation set, and the Transformer model is tested based on the test set to obtain a trained Transformer model; Load the trained Transformer model to extract knowledge from unlabeled data, post-process the entity label sequence, and obtain data in the form of entity-relationship-entity triples.

5. The method for constructing a knowledge graph for industrial robot welding process according to claim 4, characterized in that: The specific method for labeling unstructured data is as follows: Use OCR algorithm to identify text information of unstructured data and convert it into txt files; Perform data cleaning on the text to remove noise in the text; According to the welding process entities constructed in the pattern layer, the text is annotated using the BIOES annotation strategy to distinguish different welding process entities; Define the relationship between entities in the welding process and use the same annotation strategy for annotation.

6. The method for constructing a knowledge graph for industrial robot welding process according to claim 4, characterized in that: The FLAT algorithm is used to train the Transformer model. The specific method is as follows: Perform word embedding and position encoding on the input text; Train through the self-attention mechanism layer to calculate the correlation between characters and words; Perform residual connection and normalization operations on the training results to stabilize the training results; Perform nonlinear transformation on the training results to obtain transformation results; Apply addition and standardization operations to the transformation results to stabilize the output of the model; Convert the transformation result into the final output result and output the entity label sequence.

7. The method for constructing a knowledge graph for industrial robot welding process according to claim 1, characterized in that: The similarity algorithm based on edit distance is used for data alignment. A threshold is set. When the calculation result is greater than the threshold, the two entities are considered to be overlapping and then merged. The specific calculation formula is as follows: in, Indicates the first string i characters with the first character of the second string j The edit distance between characters; Represents a replacement operation. Indicates a delete operation. Represents an insert operation.

8. A system for constructing a knowledge graph for industrial robot welding process, characterized in that: include: Data collection module, used to collect industrial robot welding process data and establish an industrial robot welding process data database; The model layer construction module is used to construct the welding knowledge graph model layer from top to bottom based on the welding process data database; The data layer construction module is used to construct the knowledge graph data layer from the bottom up based on the welding knowledge graph model layer, and adopt the knowledge extraction algorithm to form the entity-relationship-entity triple data form; The data alignment module is used to align the entity-relationship-entity triple data using a similarity algorithm and merge equivalent entities and relationships. The storage and visualization module is used to store all triple data and import them into the Neo4j graph database to realize the visualization of the knowledge graph.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a method for constructing a welding process knowledge graph for an industrial robot as described in any one of claims 1 to 7 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for constructing a welding process knowledge graph for an industrial robot as described in any one of claims 1 to 7 are implemented.

Images