Runner defect intelligent diagnosis and evaluation method based on knowledge graph

Abstract

Embodiments of the present invention provide a runner defect intelligent diagnosis and evaluation method based on a knowledge graph, comprising the following steps: collecting basic information including data, pictures, standards, and regulations related to design, manufacturing, testing, acceptance, operation, and maintenance of a runner of a hydro turbine generator set; by means of analysis and screening, reviewing and retaining data having high reliability and practicality; on the basis of the data, pictures, standards, and regulations and a similar case, extracting, from the data, defect diagnosis related service knowledge as well as entities, relations, and attributes, and constructing a defect diagnosis concept graph and a defect diagnosis entity graph; and sorting defect diagnosis knowledge points, performing semantic representation on runner defect diagnosis knowledge, and establishing a diagnosis rule knowledge base. According to the present invention, a repetition semantic matching model is used to perform text and extraction rule matching for different defects so that a relation instance of defect representation features can be obtained, the defect of a runner is evaluated on the basis of the relation instance and an accurate diagnosis result is obtained, so as to guide production.

Description

An intelligent diagnosis and evaluation method for runner defects based on knowledge graph Technical Field

[0001] The present disclosure relates to the technical field of turbine runner defect diagnosis, and in particular to a knowledge graph-based intelligent runner defect diagnosis and evaluation method. Background Art

[0002] During each unit overhaul, if defects are discovered in a hydro-turbine generator unit's runner, existing technologies for diagnosing and assessing hydro-turbine generator runner defects rely on manual observation and technical analysis, combined with relevant test methods for comprehensive inference and evaluation. Knowledge graph technology is lacking in this area. Because the forms, mechanisms, and causes of hydro-turbine generator unit defects are complex, defect analysis requires a vast amount of knowledge and practical experience, making it difficult for technicians to systematically and comprehensively master them. The diverse range of procedures, specifications, and technical standards required for hydro-turbine generator unit design, operation, overhaul, and testing makes it difficult for individuals to develop a unified approach to support defect diagnosis, analysis, and evaluation. Furthermore, individual technicians have varying knowledge bases. While they have accumulated extensive work experience and data, representing a vast treasure trove of diagnostic practical knowledge, this knowledge has yet to be fully explored, consolidated, and utilized. Therefore, knowledge graph technology is needed to assist manual fault diagnosis and assessment of hydro-turbine generator unit runners.

[0003] Summary of the Invention

[0004] The embodiments of the present disclosure provide a method for intelligent diagnosis and assessment of runner defects based on a knowledge graph, the method comprising the following steps:

[0005] S1 collects basic information on data, pictures, standards and regulations related to the design, manufacture, testing, acceptance, operation and maintenance of turbine generator runners;

[0006] S2 verifies and retains data with high reliability and practicality through analysis and screening;

[0007] S3 extracts defect diagnosis-related business knowledge and entities, relationships, and attributes from the data based on the data, images, standards, procedures, and similar cases, and constructs defect diagnosis concept maps and entity maps;

[0008] S4 sorts out the defect diagnosis knowledge points, semantically represents the wheel defect diagnosis knowledge, and establishes a diagnosis rule knowledge base to provide knowledge management and service functions.

[0009] In some possible implementations of the present invention, S3 is specifically: mapping the defect records of each time period of wheel design, manufacturing, testing, acceptance, operation and maintenance and the historical records of the wheel into attention weights of wheel defects through repeated semantic matching defects, and dividing the defect types corresponding to historical omissions into different defect levels according to the attention weights, and converting them into text information; identifying entities, entity attribute information and the relationship between named entities from text information; and constructing a defect diagnosis concept map and entity map by combining defect entities, defect record entities and their entity attribute information.

[0010] In some implementations of the present invention, the repeated semantic matching is specifically: semantic pre-matching is performed on the defect records of each time period to achieve text encoding, and for a group of defect records, the combination target C = α·C is minimized. gen +β·C reg Obtain defect representation features, where α and β are factors for adjusting parameter balance and regularization; C gen The difference between the similarity value between a certain defect and a defect record and the similarity value between a certain defect and a defect record that do not match.

[0011] C gen max{0,1-sam_scri}-max{0,1-dif_scrj}; where sam_scri is the similarity value between a certain defect and a defect record, and dif_scrj is the similarity value between a certain defect and a record; C reg is the sum of the attention weights for matching a given defect Pi with defect records in all time periods, C reg γ·ΣiPi·log Reci; where γ is a weight factor used to modify the sum of the attention weights.

[0012] Compared with the prior art, the technical solution of the present invention has the following technical effects:

[0013] Attention is paid to the impact of defect records in different time periods on the final evaluation results of wheel defects. The defect records in different time periods are matched between text and extraction rules through repeated semantic matching models. Furthermore, relational instances of defect representation features are obtained. Based on the relational instances, the defects of the wheel and the treatment methods are evaluated to obtain accurate diagnosis and evaluation results to guide production.

[0014] It should be understood that the contents described in the Summary of the Invention section are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. DETAILED DESCRIPTION

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions disclosed in the present disclosure will be clearly and completely described below in conjunction with the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, but not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present disclosure.

[0016] Because the forms, mechanisms, and causes of hydro-turbine generator unit runner defects are complex, the amount of knowledge and field experience required for defect analysis is enormous, making it difficult for technicians to systematically and comprehensively master them. The various regulations, specifications, and technical standards required for the design, operation, maintenance, and testing of hydro-turbine generator units are numerous, making it difficult for individuals to form a holistic and unified approach to support their defect diagnosis, analysis, and evaluation. Individual technicians have varying knowledge reserves, and although they have accumulated a wealth of work experience and data, a vast treasure trove of diagnostic practical knowledge, this has yet to be fully mined, accumulated, and utilized. Therefore, it is necessary to introduce knowledge graph technology to assist manual fault diagnosis and evaluation of hydro-turbine generator unit runners.

[0017] The following describes in detail the method and device for obtaining the weight of dry sludge transported out provided by the embodiments of the present disclosure through specific examples.

[0018] As an implementation of the present invention, this embodiment provides a method for intelligent diagnosis and assessment of runner defects based on a knowledge graph, characterized in that the method comprises the following steps:

[0019] S1 collects basic information on data, pictures, standards and regulations related to the design, manufacture, testing, acceptance, operation and maintenance of turbine generator runners;

[0020] S2 verifies and retains data with high reliability and practicality through analysis and screening;

[0021] S3 extracts defect diagnosis-related business knowledge and entities, relationships, and attributes from the data based on the data, images, standards, procedures, and similar cases, and constructs defect diagnosis concept maps and entity maps to provide a basic cognitive layer for runner defect diagnosis.

[0022] S4 sorts out the knowledge points of defect diagnosis, semantically represents the knowledge of wheel defect diagnosis, establishes a diagnosis rule knowledge base to provide knowledge management and service functions, guides technical personnel to preliminarily determine the cause of the defect, and retrieves the recommended measures for the corresponding defect from the knowledge base, assisting technical personnel to solve, analyze, evaluate and handle defects faster and better.

[0023] In some preferred embodiments, S3 specifically includes: mapping the defect records of each time period of wheel design, manufacturing, testing, acceptance, operation and maintenance and the historical records of the wheel into attention weights of wheel defects through repeated semantic matching defects, and dividing the defect types corresponding to historical omissions into different defect levels according to the attention weights, and converting them into text information; identifying entities, entity attribute information and the relationship between named entities from text information; and constructing a defect diagnosis concept map and entity map by combining defect entities, defect record entities and their entity attribute information.

[0024] The repeated semantic matching is specifically as follows: semantic pre-matching is performed on the defect records of each time period to achieve text encoding, and for a group of defect records, the combination target C = α·C is minimized. gen +β·C reg Obtain defect representation features, where α and β are factors for adjusting parameter balance and regularization; C gen The difference between the similarity value between a certain defect and a defect record and the similarity value between a certain defect and a defect record that do not match.

[0025] C gen max{0,1-sam_scri}-max{0,1-dif_scrj}; where sam_scri is the similarity value between a certain defect and a defect record, and dif_scrj is the similarity value between a certain defect and a record; C reg is the sum of the attention weights for matching a given defect Pi with defect records in all time periods, C reg γ·ΣiPi·log Reci; where γ is a weight factor used to modify the sum of the attention weights.

[0026] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, without limitation.

[0027] The above content is only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for intelligent diagnosis and evaluation of runner defects based on knowledge graph, characterized in that: The method comprises the following steps: S1 collects basic information on data, pictures, standards and procedures for the design, manufacture, testing, acceptance, operation and maintenance of turbine generator sets; S2 analyzes and screens, proofreads and retains data with high reliability and practicality; S3 extracts defect diagnosis-related business knowledge and entities, relationships and attributes from the data based on the data, pictures, standards, procedures and similar cases, and constructs defect diagnosis concept maps and entity maps; S4 sorts out the knowledge points of defect diagnosis, semantically represents the knowledge of wheel defect diagnosis, and establishes a diagnosis rule knowledge base to provide knowledge management and service functions.

2. The method for intelligent diagnosis and evaluation of runner defects based on knowledge graph according to claim 1 is characterized in that: S3 is specifically as follows: the defect records of each time period of wheel design, manufacturing, testing, acceptance, operation and maintenance and the historical records of the wheel are mapped into the attention weights of the wheel defects through repeated semantic matching defects, and the defect types corresponding to the historical omissions are divided into different defect levels according to the attention weights, and converted into text information; entities, entity attribute information and the relationship between named entities are identified from the text information; defect diagnosis concept maps and entity maps are constructed by combining defect entities, defect record entities and their entity attribute information.

3. The method for intelligent diagnosis and evaluation of runner defects based on knowledge graph according to claim 2 is characterized in that: The repeated semantic matching is specifically as follows: semantic pre-matching is performed on the defect records in each time period to achieve text encoding, and for a group of defect records, the combination target C = α·C is minimized. gen +β·C reg The defect representation characteristics are obtained, where α and β are factors for adjusting parameter balance and regularization; C gen The difference between the similarity value between a certain defect and a defect record that matches and the similarity value between a certain defect and a defect record that does not match. C gen max{0,1-sam_scri}-max{0,1-dif_scrj}; sam_scri is the similarity value between a defect and a defect record, and dif_scrj is the similarity value between a defect and a record; C reg is the sum of the attention weights for matching a given defect Pi with defect records in all time periods, C reg γ·ΣiPi·log Reci; where γ is a weight factor used to correct the sum of the attention weights.