Intelligent auxiliary decision-making system and method for ship power system equipment fault based on knowledge graph

Abstract

The invention provides an intelligent auxiliary decision-making system and method for a ship power system equipment fault based on a knowledge graph, and belongs to the technical field of equipment operation and maintenance and industrial intelligence. On the basis of the operation and maintenance knowledge graph, an intelligent reasoning module, an intelligent decision module and an intelligent question and answer module are designed, and the accuracy of matching of equipment faults and maintenance strategies is improved. The method comprises the following steps: constructing an expert knowledge text based on maintenance regulations, historical maintenance records and expert experience of ship power system equipment; and performing data preprocessing on the operation and maintenance text data to generate an annotation text data set in the aspect of operation and maintenance of the ship power system equipment. On this basis, using CaseRel to train an initial model, performing entity recognition and relation extraction on the operation and maintenance text, constructing a head entity-relation-tail entity triple, and generating a ship power system equipment operation and maintenance knowledge graph. Based on the operation and maintenance knowledge graph, an inference rule is customized, an intelligent question and answer module is designed, intelligent auxiliary decision suggestions are displayed at the front end of the system, and development of the ship power system equipment fault intelligent auxiliary decision system is completed.

Description

Technical Field

[0001] The present invention belongs to the technical fields of equipment operation and maintenance and industrial intelligence, and particularly relates to an intelligent auxiliary decision-making system and method for equipment faults of a ship power system based on a knowledge graph. Background Art

[0002] The ship power system is a complex of all mechanical equipment and systems that ensure the normal navigation, normal operation, and berthing of a ship, and ensure the normal work, life, and safety of personnel. Its main task is to generate various energies with a certain power, realize energy conversion and distribution, so as to facilitate the normal navigation and operation of the ship. The ship power system includes a main propulsion device, an auxiliary energy supply device, equipment for ensuring the safe operation of the ship, equipment for meeting the normal life of crew members, and environmental protection equipment, etc. As the heart and artery of the entire ship, its safe and reliable operation will directly affect the safety of shipping. In recent years, both the automation level and complexity of the ship power system have been increasing continuously. However, its harsh operating conditions and strong time-variability have increased the probability of equipment failures, posing a serious threat to the life and property safety of the ship and the people on board. In the statistical data of ship failures, the failures caused by the ship power system rank first, accounting for about 60% of the total. Therefore, reducing the failure rate of the ship power system is of extremely important significance for ensuring the safe navigation of the ship.

[0003] Due to the complex types of potential safety hazards of ship power system equipment and its dynamic changes with time series, its operating conditions are changeable and the operating environment is harsh. It is difficult to predict its faults and perform maintenance in a timely manner. The maintenance means mainly rely on the experience and subjectivity of operation and maintenance personnel, with great uncertainty, leaving a large potential safety hazard for the stable operation of the equipment. Therefore, it is urgent to integrate advanced artificial intelligence technologies into the field of equipment operation and maintenance and safety management of ship power systems.

[0004] As a basic technology of artificial intelligence, the knowledge graph has extremely strong data expression ability and modeling flexibility. By modeling the association relationships between data, it can effectively organize fragmented data. With the rapid development of artificial intelligence technologies, technologies such as big data mining and knowledge extraction are widely applied to the operation and maintenance and safety management of industrial equipment. Combining advanced technologies such as knowledge graphs, intelligent question answering, intelligent reasoning, and intelligent decision-making with the on-site equipment operation and maintenance of ship power systems to develop an intelligent auxiliary decision-making system can control equipment faults from the source, improve the governance system for hidden danger investigation, improve the technical level of hidden danger rectification, and promote the development of the equipment safety governance project of ship power systems. Summary of the Invention

[0005] In view of the above technical problems, the present invention provides an intelligent auxiliary decision-making system and method for equipment faults in a ship power system based on a knowledge graph, which can quickly and accurately provide auxiliary decision-making suggestions for the operation and maintenance of key equipment in the ship power system.

[0006] To solve the above technical problems, the present invention adopts the following technical solutions.

[0007] An intelligent auxiliary decision-making system and method for equipment faults in a ship power system based on a knowledge graph includes the following steps:

[0008] S1. Based on the maintenance regulations, historical maintenance records and expert experience of the equipment in the ship power system, construct an expert knowledge text in the field of intelligent auxiliary decision-making for key equipment in the ship power system;

[0009] S2. Perform data preprocessing on the operation and maintenance knowledge text data constructed in S1 to provide a data basis for the construction of the operation and maintenance knowledge graph and the development of the intelligent auxiliary decision-making system;

[0010] S3. Based on the operation and maintenance text data provided in S2, use the CasRel entity relationship joint extraction model to separate the entities and their corresponding relationships in the operation and maintenance text, and construct an operation and maintenance knowledge graph of the equipment in the ship power system in combination with a graph database;

[0011] S4. Define intelligent inference rules, and in combination with the knowledge graph knowledge base, complete the extension and expansion of entities and relationships in the operation and maintenance text;

[0012] S5. Design an intelligent question-answering module, and in combination with the intelligent inference rules proposed in S4, complete the development of the intelligent auxiliary decision-making system for equipment faults in the ship power system.

[0013] Further, step S1 includes the following sub-steps:

[0014] S11. According to the composition structure and operation and maintenance scope of the ship power system, sort out the maintenance processes and maintenance records of relevant equipment and components;

[0015] S12. Further integrate the maintenance processes and maintenance records according to the on-site operation and maintenance expert experience, and remove the redundant parts in the operation and maintenance text;

[0016] S13. Classify the operation and maintenance text data of each equipment and component to generate expert knowledge texts related to each equipment.

[0017] Further, step S2 includes the following sub-steps:

[0018] S21. Perform data cleaning on the structured, semi-structured and unstructured data of the equipment in the ship power system, screen the text data related to the equipment, and delete the duplicate and invalid maintenance text data;

[0019] S22. Based on the text data after data cleaning in S21, different types of tags are added to the text data entities and relationships through manual annotation to construct an annotated text data set related to the intelligent decision-making of ship power system equipment.

[0020] Furthermore, step S3 includes the following sub-steps:

[0021] S31. Send the text data set constructed in S22 into the CasRel model, which first identifies all possible subjects;

[0022] S32. Under the known given category relationships, combine the subjects identified in S31 to identify the objects related to the subjects, thereby constructing triples of head entities, relationships, and tail entities.

[0023] Furthermore, step S4 includes the following sub-steps:

[0024] S41. With the knowledge base and inference engine as the core of intelligent reasoning, use the triples generated in S3 as the knowledge base for intelligent reasoning, and formulate corresponding inference rules for the inference engine;

[0025] S42. Build a data communication channel through the HTTP protocol, combine the diagnostic results of ship power system equipment, design a database query statement template, and automatically query and display the corresponding triples and inference results based on equipment failures.

[0026] Furthermore, step S5 includes the following sub-steps:

[0027] S51. Based on the intelligent reasoning module in S4, combine the graph database and the operation and maintenance knowledge graph, and set the trigger keywords for questions in the intelligent question-answering module;

[0028] S52. Provide corresponding answer templates for the questions in S51, and embed the inference results and triples in S4 into the answer templates to generate intelligent auxiliary decision-making suggestions to help on-site operation and maintenance personnel.

[0029] Adopting the present invention has the following beneficial effects: This design method aims at the problem that on-site operation and maintenance staff of ship power system equipment have difficulty in ensuring the effectiveness of fault repair. Based on the operation and maintenance knowledge graph, intelligent reasoning, intelligent decision-making, and intelligent question-answering modules are designed to improve the accuracy of matching equipment failures and repair strategies, providing an effective method and idea for enhancing the reliability of fault operation and maintenance and safety management of ship power system equipment. Brief Description of the Drawings

[0030] Figure 1 It is a flowchart of an intelligent auxiliary decision-making system and method for ship power system equipment faults based on a knowledge graph according to an embodiment of the present invention. Specific embodiments

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] As Figure 1 shown, the embodiments of this specification provide a flowchart of an intelligent auxiliary decision-making system and method for equipment faults in a ship power system based on a knowledge graph. From Figure 1 it can be seen that in one or more embodiments of this specification, an intelligent auxiliary decision-making system and method for equipment faults in a ship power system based on a knowledge graph specifically include the following steps:

[0033] S1. Based on the maintenance regulations, historical maintenance records, and expert experience of ship power system equipment, construct an expert knowledge text in the field of intelligent auxiliary decision-making for key equipment in the ship power system.

[0034] The present invention divides the units according to the system, equipment, and component levels based on the maintenance regulations, historical maintenance records, and expert experience of ship power system equipment, in combination with the scope and field of different ship power system equipment, and generates relevant expert knowledge texts.

[0035] Step S1 includes the following sub-steps:

[0036] S11. According to the composition structure and operation and maintenance scope of the ship power system, sort out the maintenance processes and maintenance records of relevant equipment and components.

[0037] The structure of the ship power system is divided into three aspects: the ship power system, ship power system equipment, and equipment components.

[0038] The operation and maintenance text includes various types of entities such as system names, equipment names, component names, composition relationships, failure causes, and maintenance suggestions.

[0039] Based on the structure divided by the ship power system, sort out the relevant maintenance processes and maintenance records at different levels of the unit.

[0040] S12. According to the experience of on-site operation and maintenance experts, further integrate the maintenance processes and maintenance records, and remove the redundant parts in the operation and maintenance text.

[0041] Define and constrain different structures of the unit, retain the knowledge data of the relevant parts in the operation and maintenance text, and perform operations of deletion, modification, and checking on the irrelevant parts.

[0042] S13. Classify the operation and maintenance text data of each part to generate corresponding expert knowledge text.

[0043] S2. Perform data preprocessing on the operation and maintenance knowledge text data constructed in S1 to provide a data basis for the construction of the operation and maintenance knowledge graph and the development of the intelligent auxiliary decision-making system.

[0044] Step S2 includes the following sub-steps:

[0045] S21. Perform data cleaning on the structured, semi-structured, and unstructured data of the ship power system equipment, screen the text data of the equipment involved in the system, and delete the duplicate and invalid maintenance text data.

[0046] In the embodiments of this specification, obtain the overhaul regulations and historical maintenance record text data and data formats of the ship power system equipment, classify the text data according to different data formats, uniformly merge the data of different formats, and modify or delete the data with incorrect formats.

[0047] S22. Based on the text data after data cleaning in S21, add different types of labels to the text data entities and relationships through manual annotation to construct a text data set with labels related to intelligent decision-making of ship power system equipment.

[0048] After manual processing, a total of 3 relationship types, namely "causes", "suggestions", and "constitutes", are included in all text data. Process the relationship types into the JSON format required for model training. According to the expert knowledge contained in the auxiliary decision-making text data of the ship power system equipment, divide the entities connected to the 3 relationship types into 6 entity types, namely "system", "equipment failure", "component failure", "maintenance method", "cause", and "maintenance suggestion".

[0049] S3. Based on the operation and maintenance text data provided in S2, use the CasRel entity relationship joint extraction model to extract the entities and their corresponding relationships in the operation and maintenance text, and construct an operation and maintenance knowledge graph of ship power system equipment in combination with a graph database.

[0050] Pre-train a joint extraction model for operation and maintenance entity relationships of ship power system equipment based on CasRel, input the operation and maintenance text data generated in S2 into the model, construct triples of head entity-relationship-tail entity, and generate an operation and maintenance knowledge graph of ship power system equipment in combination with the Neo4j graph database.

[0051] Step S3 includes the following sub-steps:

[0052] S31. Send the text data set constructed in S22 into the CasRel model, and the model first identifies all possible subjects.

[0053] Construct an entity recognition module using a linear layer and a Sigmoid activation function. Through this module, determine the start and end positions of the head entity to identify the head entity in the text.

[0054] Using the nearest matching principle and combining with the entity recognition module, pair the identified start and end positions to obtain a set of candidate head entities.

[0055] S32. Under the known given category relationships, combine with the subject recognized in S31 to identify the object related to the subject, thereby constructing a triple of head entity, relationship, and tail entity.

[0056] Perform joint recognition of relationships and tail entities, mainly through a set of tail entity recognition layers related to relationships.

[0057] Among them, the tail entity recognition layer has the same structure as the head entity recognition layer, but the input considers not only the output of the text encoding layer but also the features of the head entity.

[0058] Take the relationship as prior knowledge, respectively identify the head entity and the tail entity, thereby forming a triple of head entity - relationship - tail entity, and combine with the Neo4j database to generate a knowledge graph of ship power system equipment operation and maintenance.

[0059] S4. Customize intelligent reasoning rules, combine with the knowledge graph knowledge base, and complete the extension and expansion of entities and relationships in the operation and maintenance text.

[0060] Step S4 includes the following sub - steps:

[0061] S41. Take the knowledge base and the inference engine as the core of intelligent reasoning. Use the triples generated in S3 as the knowledge base for intelligent reasoning, and formulate corresponding inference rules for the inference engine.

[0062] The intelligent reasoning module is mainly composed of a knowledge base and an inference engine. Among them, the knowledge base takes Jena as the core, and the inference engine takes Fuseki as the core.

[0063] Use a Python script to import the triples of the ship power system operation and maintenance knowledge graph in the Neo4j graph database into Jena as the knowledge base required for knowledge reasoning.

[0064] Combine with the structure composition of the ship power system equipment to customize inference rules, and write the inference rules into Fuseki to construct an inference engine.

[0065] S42. Construct a data communication channel through the HTTP protocol. Combine with the ship power system equipment diagnosis results, design a database query statement template, and automatically query and display the corresponding triples and inference results based on equipment failures.

[0066] Build a communication channel through the HTTP protocol. Combine the fault diagnosis results of the ship power system equipment, design a Sparql query template for the front-end request. When the system diagnoses a fault in the ship power system equipment, the program reads the fault nodes and generates the corresponding Sparql query statement, and sends a request to the Fuseki inference engine server. After receiving the request, Fuseki performs queries and inferences based on the inference rules in the triple data support provided by Jena.

[0067] S5. Design an intelligent question-answering module. Combine the intelligent inference rules proposed in S4 to complete the development of the intelligent auxiliary decision-making system for ship power system equipment faults.

[0068] S51. Based on the intelligent inference module in S4, combine the graph database and the operation and maintenance knowledge graph, and set the trigger keywords for the questions in the intelligent question-answering module.

[0069] Classify the questions in the intelligent question module into three types: "fault cause", "maintenance suggestion", and "unit structure composition", and set the corresponding trigger keywords for each of the three types of questions.

[0070] S52. Provide the corresponding answer templates for the questions in S51, and embed the inference results and triples in S4 into the answer templates to generate intelligent auxiliary decision-making suggestions to help on-site operation and maintenance personnel.

[0071] The question-and-answer results can be returned through the HTTP protocol and displayed correspondingly at the front end of the system for operation and maintenance personnel to view and assist in making relevant decisions and maintenance.

[0072] This design method aims at the problem that the decision-making for the fault repair of ship power system equipment depends on manual experience and is interfered by the subjective factors of operation and maintenance personnel. By constructing an intelligent auxiliary decision-making system and method for ship power system equipment faults based on the knowledge graph, it provides auxiliary decision-making suggestions for the generation of maintenance plans for operation and maintenance personnel, effectively enabling the operation and maintenance personnel to make decisions without being interfered by subjective factors, and providing an effective method and idea for improving the reliability of the operation and maintenance of ship power system equipment.

[0073] It should be understood that the exemplary embodiments described herein are illustrative and not restrictive. Although one or more embodiments of the present invention have been described in conjunction with the accompanying drawings, those of ordinary skill in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the present invention as defined by the appended claims.

Claims

1. A knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: The following steps are involved: S1. Based on the maintenance procedures, historical maintenance records and expert experience of ship power system equipment, construct expert knowledge text in the field of intelligent auxiliary decision-making of key equipment of ship power system; S2, pre-process the operation and maintenance knowledge text data constructed in S1 to provide a data basis for the construction of the operation and maintenance knowledge graph and the development of the intelligent auxiliary decision-making system; S3, based on the operation and maintenance text data provided by S2, uses the CasRel entity-relationship joint extraction model to separate the entities and entity-corresponding relationships in the operation and maintenance text, and builds a knowledge graph for the operation and maintenance of ship power system equipment in combination with the graph database; S4. Customize intelligent reasoning rules and combine with knowledge graph knowledge base to complete the extension and expansion of entities and relationships in operation and maintenance texts; S5. Design an intelligent question-answering module and combine it with the intelligent reasoning rules proposed in S4 to complete the development of an intelligent auxiliary decision-making system for ship power system equipment failures.

2. According to claim 1, a knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: Step S1 includes the following sub-steps: S11. Organize the inspection and maintenance process and maintenance records of relevant equipment and components according to the composition structure and operation and maintenance scope of the ship power system; S12. Based on the experience of on-site operation and maintenance experts, further integrate the maintenance process and maintenance records to remove redundant parts in the operation and maintenance text; S13. Classify the operation and maintenance text data of each device and component, and generate expert knowledge text related to each device.

3. According to claim 1, a knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: Step S2 includes the following sub-steps: S21. Perform data cleaning on the structured, semi-structured and unstructured data of ship power system equipment, screen the text data related to the equipment, and delete duplicate and invalid maintenance text data; S22. Based on the text data cleaned by S21 data, different types of labels are added to the text data entities and relationships through manual annotation methods to construct an annotated text dataset related to intelligent decision-making of ship power system equipment.

4. According to claim 1, a knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: Step S3 includes the following sub-steps: S31, the text dataset constructed in S22 is fed into the CasRel model, which first identifies all possible subjects; S32. Under a known given category relationship, combined with the subject identified in S31, identify the object related to the subject, thereby constructing a triple of head entity, relationship, and tail entity.

5. According to claim 1, a knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: Step S4 includes the following sub-steps: S41, taking the knowledge base and inference engine as the core of intelligent reasoning, the triples generated by S3 are used as the knowledge base of intelligent reasoning, and corresponding inference rules are formulated for the inference engine; S42. Build a data communication channel through the HTTP protocol, combine the diagnostic results of the ship power system equipment, design a database query statement template, and automatically query and display the corresponding triples and reasoning results based on the type of equipment failure.

6. According to claim 1, a knowledge graph-based intelligent auxiliary decision-making system and method for ship power system equipment failure, characterized in that: Step S5 includes the following sub-steps: S51, based on the intelligent reasoning module of S4, combined with the graph database and the operation and maintenance knowledge graph, set the trigger keywords for the questions in the intelligent question and answer module; S52: Provide a corresponding answer template for question S51, and embed the reasoning results and triples of S4 into the answer template to generate intelligent decision-making suggestions to provide assistance to on-site operation and maintenance personnel.

Images