Ternary data knowledge extraction model and ternary data fusion method for equipment abnormality diagnosis and maintenance

Abstract

A ternary data fusion method for equipment abnormality diagnosis and maintenance includes the following steps: Step 1: Based on the characteristics of the field of equipment operation and maintenance, a specific definition of human-machine-object ternary data is made, including the source, characteristics and internal connections between the ternary data; Step 2: An ontology model is constructed in a top-down manner; Step 3: Based on the ontology model, a joint entity and relationship extraction task is performed, and a ternary data knowledge extraction model is used to extract all triples from sentences; Step 4: Based on the triples, a knowledge graph of human-machine-object for equipment operation and maintenance is constructed to achieve the fusion of human-machine-object ternary data. The ternary data knowledge extraction model and ternary data fusion method for equipment abnormality diagnosis and maintenance of the present invention designs and develops an equipment operation and maintenance human-machine-object ontology based on the data characteristics of diversity, multi-source and heterogeneity of human-machine-object ternary data, and simultaneously uses the ternary data knowledge extraction model to extract triples to better utilize the text information of related processes.

Description

Technical Field

[0001] The present invention belongs to the technical field of abnormal diagnosis and maintenance of equipment failures, and specifically provides a ternary data knowledge extraction model and a ternary data fusion method for equipment abnormal diagnosis and maintenance. Background Art

[0002] In industrial settings, the operating status of equipment directly impacts product quality, production efficiency, and safety. Equipment status monitoring, anomaly diagnosis, defect identification, and operation and maintenance decision-making are key to reducing unplanned downtime, extending equipment life, and ensuring production efficiency, and are therefore of great research significance and practical value. Currently, research in the field of equipment operation and maintenance can be primarily categorized into two areas: data-driven equipment health status analysis and knowledge-driven intelligent equipment operation and maintenance decision-making. Data-driven equipment status analysis involves analyzing, classifying, and predicting monitoring signals based on statistical analysis, machine learning, deep learning, and other methods, thereby enabling monitoring and diagnosis of equipment operating status and faults. Knowledge-driven operation and maintenance decision-making utilizes knowledge engineering techniques to standardize and reuse expert operation and maintenance experience. By building knowledge bases and expert systems, it enables intelligent search, intelligent question-answering, and intelligent decision-making for equipment operation and maintenance solutions. In practical engineering applications, data-driven approaches offer high accuracy, but ensuring accurate results often requires high-level information processing skills from field workers, requiring human experience. Knowledge-driven approaches, while more versatile, often fail to provide accurate diagnostic results based solely on empirical knowledge, requiring the integration of data analysis. Therefore, the operation and maintenance problems of complex equipment are difficult to be effectively solved simply by relying on expert experience or model diagnosis. Researching knowledge and data fusion-driven models is the key to improving the practical engineering application value of intelligent operation and maintenance methods.

[0003] In the field of equipment operation and maintenance, "human" data primarily refers to the accumulated expert experience and knowledge gained from solving practical problems, primarily contained in various troubleshooting manuals, equipment inspection records, and repair reports. "Machine" data primarily refers to the computer algorithm models used for equipment status analysis, identification, and diagnosis, as well as the data features extracted from them. "Object" data primarily refers to on-site records representing equipment status, equipment signal data collected by various sensors, and attribute information of various equipment. The "human," "machine," and "object" ternary data are multi-source, heterogeneous data characterized by large data volumes, uneven data quality, and mixed and discrete data. Fusion of these ternary data is a significant challenge. Currently, multivariate data fusion can be achieved using methods based on statistics, estimation, evidence theory, and information theory. While these methods aggregate, optimize, and redistribute various data sources, they cannot effectively analyze correlations and establish dependencies between heterogeneous data.

[0004] As a relational semantic network, the knowledge graph can establish connections between data, enabling data integration and connectivity. Furthermore, the knowledge graph can provide deep relationship discovery and reasoning capabilities for decision support. A key step in building a knowledge graph is designing an appropriate ontology structure. Another key issue is how to better utilize the textual information of relevant processes, specifically, effectively extracting relationship triples. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a ternary data knowledge extraction model and ternary data fusion method for equipment abnormality diagnosis and maintenance. According to the data characteristics of diversity, multi-source and heterogeneity of human-machine-object ternary data, an equipment operation and maintenance human-machine-object ontology is designed and developed. At the same time, the ternary data knowledge extraction model is used to extract triples to better utilize the text information of related processes.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention first proposes a ternary data knowledge extraction model for equipment abnormality diagnosis and maintenance, including:

[0008] Embedding layer, which is used to encode the input sentence and convert it into a sentence vector;

[0009] An encoding layer, which jointly encodes words based on the influence of context word information to extract text sequence features from the sentence vector to obtain a forward input vector;

[0010] A hybrid attention layer performs feature weighting on the forward input vector to obtain a forward feature vector to improve the representation ability of the vector;

[0011] The s_model module includes a first convolutional layer and a first decoding layer, wherein the first convolutional layer is used to reduce the size of the forward feature vector and the first decoding layer is used to identify all potential subjects in the sentence;

[0012] The po_model module includes a fusion layer, a second convolutional layer, and a second decoding layer. The fusion layer is used to perform feature fusion on a group of subject entities randomly sampled by the first decoding layer with the forward feature vector to obtain a feature fusion vector. After the feature fusion vector passes through the second convolutional layer, the second decoding layer identifies the object related to the subject under a given category relationship and finally outputs a triple.

[0013] Furthermore, the encoding layer adopts a bidirectional LSTM encoding layer.

[0014] Furthermore, local attention and linear attention are used jointly in the hybrid attention layer.

[0015] The present invention also proposes a ternary data fusion method for equipment abnormality diagnosis and maintenance, comprising the following steps:

[0016] Step 1: Definition of ternary data: Based on the characteristics of the equipment operation and maintenance field, the human-machine-object ternary data is specifically defined, including the source and characteristics of the data and the internal connection between the ternary data;

[0017] Step 2: Ontology model construction: Construct the ontology model in a top-down manner;

[0018] Step 3: Perform joint entity and relationship extraction based on the ontology model, and use the triple data knowledge extraction model described above to extract all triples from the sentence;

[0019] Step 4: Build a human-machine-object knowledge graph for equipment operation and maintenance based on triples to achieve the fusion of human-machine-object triple data.

[0020] Furthermore, in step 1, the method for defining the human-machine-object ternary data is:

[0021] Human data: This includes expert knowledge or experience, as well as practical experience. Expert knowledge or experience is accumulated by on-site workers or domain experts during equipment maintenance and is primarily stored in databases or documents in text form. Practical experience is accumulated by professionals using computer models to diagnose equipment failures and is primarily reflected in the use of algorithmic models.

[0022] Machine data: includes algorithms, models and extracted data features for equipment status analysis, identification and diagnosis obtained through computer modeling and training. They are mainly stored in computers and are characterized by reproducibility, rigor and operability.

[0023] Object data: includes records representing the status of equipment during operation, equipment signal data collected by various sensors, and attribute information of various production equipment. It is mainly stored in the form of text, pictures, signals, and videos, and is characterized by diversity, timeliness, and objectivity.

[0024] Furthermore, in step 2, the method for constructing the ontology model includes the following steps:

[0025] 21) Conduct industry research to determine the application and demand scope of the equipment operation and maintenance field;

[0026] 22) Define the ontology of equipment operation and maintenance, determine the architectural foundation of the ontology, including the division of human, machine and object areas, and determine the ontology structure and relationship definition;

[0027] 23) Extract and analyze the human-machine-object ontology concepts, obtain the ontology concept model, and express it using ER diagram;

[0028] 24) Carry out formal description of human-machine-object ontology, define object attributes and data attributes of ontology based on OWL, and use protégé to visualize the ontology;

[0029] 25) Conduct final verification of the ontology by experts in related fields and entity testing: if it passes the verification, the ontology model is obtained; if it fails the verification, execute step 23).

[0030] The beneficial effects of the present invention are:

[0031] The present invention proposes a ternary data fusion method for equipment abnormality diagnosis and maintenance. The constructed equipment operation and maintenance human-machine-object knowledge graph combines fault diagnosis and operation and maintenance decision-making technologies. When an equipment failure occurs, on the one hand, the fault location and cause can be diagnosed through big data analysis technology, and on the other hand, an operation and maintenance plan can be proposed through human experience and knowledge, thereby improving the comprehensive support of the technology. At the same time, the knowledge graph integrates the "human", "machine" and "object" ternary data, incorporating the previous experience of professionals in data selection, model debugging, parameter setting, etc., and can realize intelligent decision-making according to actual conditions, solving the problem of on-site workers not being able to select or use models. The knowledge graph contains a large amount of knowledge such as fault component information, fault handling measures and fault prevention measures. Through the application of the graph, rapid fault response and automatic recommendation of fault handling plans can be achieved, greatly reducing the dependence on operators and improving the quality and efficiency of equipment operation and maintenance. Compared with traditional operation and maintenance technologies, the human-machine-object knowledge graph reduces the technical requirements of on-site personnel in signal analysis during the equipment operation and maintenance process. Through the application of the graph, automatic recommendation of intelligent operation and maintenance decision-making plans can be realized, providing support for intelligent and accurate decision-making in the field of equipment operation and maintenance. In addition, the knowledge graph has the characteristics of self-updating and self-iteration, which can complete the processing and storage of new knowledge on site, achieve the purpose of coordinating the graph knowledge coverage domain with the actual production situation, and better provide fault diagnosis and operation and maintenance decision-making services for actual production. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and beneficial effects of the present invention more clear, the present invention provides the following drawings for illustration:

[0033] Figure 1 Schematic diagram of the ternary data fusion method for equipment abnormality diagnosis and maintenance of the present invention;

[0034] Figure 2 A schematic diagram of the principle of the source and internal connection of the three-dimensional data of human, machine and object;

[0035] Figure 3 Flowchart constructed for the ontology model;

[0036] Figure 4This is the structural principle diagram of the ternary data knowledge extraction model;

[0037] Figure 5 This is a schematic diagram of the human-machine-object ontology ER model in equipment operation and maintenance in a welding workshop;

[0038] Figure 6 A diagram showing the human-machine-object entity of equipment operation and maintenance;

[0039] Figure 7 The bar chart shows the comparison results of different models;

[0040] Figure 8 A visual diagram of the human-machine-object knowledge graph. DETAILED DESCRIPTION

[0041] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0042] like Figure 1 As shown, the ternary data fusion method for equipment abnormality diagnosis and maintenance in this embodiment includes the following steps:

[0043] Step 1: Definition of ternary data: Based on the characteristics of the field of equipment operation and maintenance, the human-machine-object ternary data is specifically defined, including the source, characteristics and internal connection between the ternary data, which will help further carry out the integration of ternary data.

[0044] Specifically, in this embodiment, the method for defining the human-machine-object ternary data is as follows:

[0045] Human data: This includes expert knowledge or experience and practical experience. The accumulated expert knowledge or experience of field workers or domain experts in equipment fault diagnosis, handling, and prevention during equipment maintenance is primarily stored in textual form in databases or documents, such as fault manuals, equipment inspection records, and repair reports. The practical experience accumulated by professionals when diagnosing equipment faults using computer models, including model selection, algorithm design, and parameter setting, is primarily reflected in the use of algorithmic models and possesses strong relevance, reasonability, and enforceability.

[0046] Machine data: including algorithms, models and extracted data features used for equipment status analysis, identification and diagnosis obtained through computer modeling and training. It is mainly stored in the computer in file formats such as .ckpt, .index, .pth, .data, etc., and has the characteristics of reproducibility, rigor and operability.

[0047] Object data: records representing the status of equipment during operation, equipment signal data collected by various sensors, attribute information of various production equipment, etc., mainly stored in the form of text, pictures, signals, videos, etc., with the characteristics of diversity, timeliness, and objectivity.

[0048] Step 2: Ontology model construction: Construct the ontology model in a top-down manner.

[0049] The human-machine-object ontology model is constructed to divide data sources into three categories: human knowledge data, machine algorithm data, and object signal data. According to the definition of human-machine-object data, human knowledge data mainly refers to expert knowledge and experience in the form of text, machine algorithm data mainly refers to model information used for equipment status diagnosis, and object signal data mainly refers to the collected data input into the model and equipment data, such as Figure 2 Secondly, although the human-machine-object ternary data are independent of different systems, there are also certain internal connections, such as Figure 2 As shown in the figure, "people" need to analyze and judge the training results of the "machine" to determine the specific fault; "things" serve as data input for the "machine" to support its model training; "people" collect and record "things"; "things" and "machines" assist "people" in making decisions, etc.

[0050] The ontology design work stipulates the specific process of ontology construction and provides detailed guidance for its construction. The TOVE method, skeleton method and seven-step method are commonly used ontology construction methods, which are all extracted from practical scenarios. The equipment operation and maintenance human-machine-object ternary data ontology has the characteristics of many professional terms, data redundancy, complex attribute relationships, and few reusable ontologies. This embodiment refers to the seven-step method and adopts a top-down approach to ontology construction, aiming to start from the top-level concepts of the field, and gradually refine and expand until a complete and accurate ontology model is constructed. Specifically, Figure 3 As shown, in this embodiment, the method for constructing the ontology model includes the following steps:

[0051] 21) Conduct industry research to determine the application and demand scope of the equipment operation and maintenance field;

[0052] 22) Define the ontology of equipment operation and maintenance, determine the architectural foundation of the ontology, including the division of human, machine and object areas, and determine the ontology structure and relationship definition;

[0053] 23) Then collect Figure 2 The three-dimensional data of "man", "machine" and "object" in the text are used to extract and analyze the human-machine-object ontology concept, obtain the ontology concept model, and express it with an ER diagram;

[0054] 24) Carry out formal description of human-machine-object ontology, define object attributes and data attributes of ontology based on OWL, and use protégé to visualize the ontology;

[0055] 25) Conduct final verification of the ontology by experts in related fields and entity testing: if it passes the verification, the ontology model is obtained; if it fails the verification, execute step 23).

[0056] Step 3: Perform joint entity and relationship extraction tasks based on the ontology model, and use the triple data knowledge extraction model described above to extract all triples from the sentence.

[0057] The task of joint entity and relationship extraction based on the ontology model is aimed at extracting all relation triplets from sentences, that is, triples, the basic building blocks of the graph. This embodiment designs a joint extraction model MAREL to extract triple relationships in sentences. MAREL uses BiLSTM for joint annotation to encode sentence features, and decodes sentence features through the subject prediction module and the object prediction module under a specific relationship to extract the triple relationships in the sentence. Among them, MAREL integrates the subject features in the object prediction module to solve the common entity overlap problem in the joint extraction task. In addition, in view of the complex and lengthy characteristics of text data in the field of equipment operation and maintenance, a hybrid attention layer is added to the MAREL framework to improve the model's ability to model long texts. Specifically, in the study of the joint extraction of entities and relationships of unstructured data, this embodiment adopts an end-to-end hybrid neural network model to extract triples. The framework of the model is as follows Figure 4 As shown, the main components of the model are a subject extraction module (s_model) and a relation-specific object extraction module (po_model). s_model identifies all potential subjects in a sentence; po_model, given a categorical relation, identifies objects related to the subject. This modeling approach extracts all triples from a sentence and addresses potential entity overlap.

[0058] Specifically, the ternary data knowledge extraction model for equipment abnormality diagnosis and maintenance in this embodiment includes:

[0059] Embedding layer, which is used to encode the input sentence and convert it into a sentence vector;

[0060] The encoding layer jointly encodes the words based on the influence of contextual word information to extract text sequence features from the sentence vector and obtain a forward input vector. The encoding layer of this embodiment adopts a bidirectional LSTM encoding layer, which uses two parallel LSTM layers to jointly encode the words based on the influence of contextual word information to extract text sequence features.

[0061] The hybrid attention layer performs feature weighting on the forward input vector to obtain a forward feature vector, implicitly discarding some noise corpus to improve the vector's representational ability. The hybrid attention layer of this embodiment combines local attention with linear attention.

[0062] The s_model module includes the first convolutional layer and the first decoding layer. The first convolutional layer is used to reduce the size of the forward feature vector, and the first decoding layer is used to identify all potential subjects in the sentence.

[0063] The po_model module includes a fusion layer, a second convolutional layer, and a second decoding layer. The fusion layer is used to fuse a set of subject entities (k1, k2) randomly sampled by the first decoding layer with the forward feature vector to obtain a feature fusion vector. After the feature fusion vector passes through the second convolutional layer, the second decoding layer identifies the object related to the subject under a given category relationship and finally outputs a triplet (S, P, O).

[0064] Step 4: Store triple data in the Neo4j database, build a human-machine-object knowledge graph for equipment operation and maintenance based on the triples, and build a human-machine-object knowledge graph for equipment operation and maintenance in a visual way to achieve the fusion of human-machine-object triple data.

[0065] The following describes in detail the ternary data fusion method for equipment abnormality diagnosis and maintenance in this embodiment with reference to specific examples.

[0066] 1. Scenario Introduction

[0067] A case study was conducted on a welding workshop in a production facility to construct a human-machine-object knowledge graph for equipment operation and maintenance. The welding workshop boasts a high level of automation and intelligence. Robots handle all processes, from material transportation and welding to gluing and product quality inspection. The workshop boasts 350 intelligent robots, along with a wide range of equipment, including welding robots, gluing machines, and hemming membranes. A survey of equipment failures in this workshop revealed a high probability of equipment failure and a high recurrence rate, ranging from 20% to 40%. However, when equipment failures occur, on-site operators must search through extensive documentation, including but not limited to equipment manuals, repair manuals, and maintenance reports, to make maintenance decisions based on the information contained within. However, this decision-making approach is inefficient, often requiring significant time to identify appropriate solutions from numerous text files. Furthermore, the decision support system and big data analytics technology are isolated, requiring experienced operations and maintenance experts to connect the two, severely impacting the efficiency and quality of equipment status management. Therefore, based on the research method of the present invention, with the support of multimodal information data of welding workshop equipment maintenance, including robot user manuals, welding workshop equipment maintenance records, fault repair 8D reports, sensor collection data and equipment fault diagnosis model data, a human-machine-object knowledge graph for equipment operation and maintenance is constructed to verify the effectiveness of the proposed method and its practical application value in engineering.

[0068] 2. Ontology display

[0069] Taking the three-dimensional data of equipment operation and maintenance in the welding workshop as the data source, the ontology is constructed based on the human-machine-object ontology construction method. First, term extraction and concept extraction are performed from the three-dimensional data of "human", "machine" and "object", and the knowledge concept of the ontology is expanded to obtain the human-machine-object ontology ER model, such as Figure 5 The tree structure shown in Figure 1 is shown in Figure 2. This modeling approach breaks down the top-level class concept into subclasses, all the way down to the bottom level. This classification method also implies the hierarchical relationship between different classes. Furthermore, subclass ontologies must also be classified according to the definition of the human-machine-object ontology. Parent and child classes with a containment relationship may belong to different human-machine-object ontology categories, and sibling classes at the same level may also come from different categories. This indicates that this ontology model initially realizes the interaction between the human-machine-object ternary, laying the foundation for ultimate integration.

[0070] After defining the classes and the hierarchical relationships between them, the next step is to define the attributes of the classes, namely object attributes and data attributes. This process of defining classes and creating attributes is called ontology representation. Using the ontology modeling tool Protégé and the OWL language to define classes and create attributes to build an ontology, and then perform the final ontology visualization, such as Figure 6 shown.

[0071] 3. Algorithm comparison

[0072] In the case study of this embodiment, the detailed information of the triple relationship dataset extracted for automated graph construction is shown in Table 1. 36 relationship types are predefined based on the content characteristics of the dataset. To verify the superior performance of the MAREL model in processing knowledge extraction tasks in the field of equipment operation and maintenance, we conducted an algorithm comparison experiment. The comparison models are FETI and CaSRel, which currently perform best in joint extraction tasks. The comparison results are shown in the figure below. Figure 7 The results show that MAREL is ahead of other models in this case dataset.

[0073] Table 1 Datasets related to equipment operation and maintenance in welding workshops

[0074]

[0075]

[0076] 4. Display of ternary data fusion graph

[0077] In the case of this embodiment, the main focus is on the construction of a graph for the operation and maintenance of important production equipment in automobile production welding workshops, such as spot welding robots, visual robots, arc welding robots, gripper robots, gluing robots, stud robots, lettering robots, and detection robots. First, under the guidance of the human-machine-object ontology model of equipment operation and maintenance, the graph is constructed to complete the mapping of some ontology to entities. Secondly, based on the MAREL model, the triple relationship of non-structured data is automatically extracted. Finally, the triple data is stored in the Neo4j database, and the human-machine-object knowledge graph of equipment operation and maintenance cases is constructed in a visual way. Some of the graph contents are as follows: Figure 8 As shown in the figure, a large amount of complex knowledge and relationships are interconnected to achieve the fusion of ternary data, where the blue part is the knowledge of "people", the red part is the knowledge of "machines", and the green part is the knowledge of "things".

[0078] The above embodiments are merely preferred embodiments for the purpose of fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are within the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A ternary data fusion method for equipment abnormality diagnosis and maintenance, characterized by: The steps include: Step 1: Definition of ternary data: Based on the characteristics of the equipment operation and maintenance field, the human-machine-object ternary data is specifically defined, including the source and characteristics of the data and the internal connection between the ternary data; Step 2: Ontology model construction: Construct the ontology model in a top-down manner; Step 3: Perform joint entity and relationship extraction based on the ontology model, and use the triple data knowledge extraction model to extract all triples from the sentence; Step 4: Build a human-machine-object knowledge graph for equipment operation and maintenance based on the triples to achieve the fusion of human-machine-object triple data; In step 3, the ternary data knowledge extraction model for equipment abnormality diagnosis and maintenance includes: Embedding layer, which is used to encode the input sentence and convert it into a sentence vector; An encoding layer, which jointly encodes words based on the influence of context word information to extract text sequence features from the sentence vector to obtain a forward input vector; A hybrid attention layer performs feature weighting on the forward input vector to obtain a forward feature vector to improve the representation ability of the vector; The s_model module includes a first convolutional layer and a first decoding layer, wherein the first convolutional layer is used to reduce the size of the forward feature vector and the first decoding layer is used to identify all potential subjects in the sentence; The po_model module includes a fusion layer, a second convolutional layer, and a second decoding layer. The fusion layer is used to perform feature fusion on a group of subject entities randomly sampled by the first decoding layer with the forward feature vector to obtain a feature fusion vector. After the feature fusion vector passes through the second convolutional layer, the second decoding layer identifies the object related to the subject under a given category relationship and finally outputs a triple.

2. The ternary data fusion method for equipment abnormality diagnosis and maintenance according to claim 1 is characterized by: The encoding layer adopts a bidirectional LSTM encoding layer.

3. The ternary data fusion method for equipment abnormality diagnosis and maintenance according to claim 1 is characterized by: The hybrid attention layer combines local attention with linear attention.

4. The ternary data fusion method for equipment abnormality diagnosis and maintenance according to claim 1 is characterized by: In step 1, the method for defining the human-machine-object triple data is: Human data: This includes expert knowledge or experience, as well as practical experience. Expert knowledge or experience is accumulated by on-site workers or domain experts during equipment maintenance and is primarily stored in databases or documents in text form. Practical experience is accumulated by professionals using computer models to diagnose equipment failures and is primarily reflected in the use of algorithmic models. Machine data: includes algorithms, models and extracted data features for equipment status analysis, identification and diagnosis obtained through computer modeling and training. They are mainly stored in computers and are characterized by reproducibility, rigor and operability. Object data: includes records representing the status of equipment during operation, equipment signal data collected by various sensors, and attribute information of various production equipment. It is mainly stored in the form of text, pictures, signals, and videos, and is characterized by diversity, timeliness, and objectivity.

5. The ternary data fusion method for equipment abnormality diagnosis and maintenance according to claim 1 is characterized by: In the second step, the method for constructing the ontology model includes the following steps: 21) Conduct industry research to determine the application and demand scope of the equipment operation and maintenance domain; 22) Define the ontology of equipment operation and maintenance, determine the architectural foundation of the ontology, including the division of human, machine and object areas, and determine the ontology structure and relationship definition; 23) Extract and analyze the human-machine-object ontology concepts, obtain the ontology concept model, and express it using ER diagram; 24) Formal description of human-machine-object ontology, definition of object attributes and data attributes of ontology based on OWL, and visualization of ontology by protégé; 25) Conduct final verification of the ontology by experts in related fields and entity testing: If it passes the verification, the ontology model is obtained; if it fails the verification, execute step 23).

Images