A text-based method for analyzing accident causes
Through the text-based accident cause analysis method, accident and hidden danger entities are extracted, and word vector algorithms and knowledge graphs are combined to build a more accurate fault tree, which solves the problem of insufficient and complete fault tree analysis in the existing technology and improves accident analysis accuracy and safety.
Patent Information
- Application Number
- CN202011615208.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-12-30
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2040-12-30
AI Technical Summary
The accuracy of fault tree analysis in the prior art in production safety depends on human experience, which leads to inadequacy and completeness of the analysis results, which may lead to the recurrence of the same type of safety accidents.
Using text-based accident cause analysis method, by extracting accident entities and hidden danger entities, using word vector algorithms and knowledge graphs to analyze fault tree, obtain accident cause analysis fault tree, and combine enterprise safety management systems and industry best practice safety standards to build More accurate failure tree.
It improves the accuracy and analysis of the fault tree, reduces the probability of similar safety accidents, and ensures the integrity and reliability of the cause of the fault.
Smart Images

Figure CN114691876B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of production safety management, and in particular to a text-based accident cause analysis method. Background Art
[0002] Fault tree analysis is an important method for analyzing system reliability and safety, and is widely used in fault diagnosis. Based on the hierarchical characteristics of faults, the relationship between the cause and consequence of a fault often has many levels and forms a series of cause-and-effect chains, and a fault tree is formed when one cause has multiple effects or one effect has multiple causes.
[0003] At present, in production safety applications, fault trees are usually established manually based on experience, and then the established fault trees are used for analysis of faults that occur during the production process. For example, the Chinese invention patent "Method and system for diagnosing faults of wind turbine generator sets" with authorization announcement number CN107544462B (application number 201710800858.X) discloses a scheme that uses an established fault tree for fault analysis, specifically determining a fault tree corresponding to the fault based on fault-related information; obtaining a fault information keyword identification group corresponding to the fault tree based on the determined fault tree; obtaining specific values of variables in the fault information keyword identification group from the fault record file; based on the specific values of the variables in the acquired fault information keyword identification group and the logical relationship between the nodes of the fault tree, searching the nodes of the fault tree to determine the cause of the fault. The accuracy of this fault analysis result depends on the accuracy of the establishment of the fault tree, and the accuracy of the fault tree established manually based on experience depends on the sufficiency and completeness of the human cause analysis. The fault trees constructed by different people are also completely different, and the corresponding analysis results may also be completely different. As a result, the fault analysis conducted through this fault tree cannot guarantee the adequacy and completeness of the analysis results, which may be an important reason for the repeated occurrence of similar safety accidents. Summary of the invention
[0004] The technical problem to be solved by the present invention is to provide a text-based accident cause analysis method for the above-mentioned prior art, which can automatically, fully and completely analyze the cause of the accident to form a more accurate fault tree.
[0005] The technical solution adopted by the present invention to solve the above-mentioned problem is: a text-based accident cause analysis method, characterized in that: accident entities are extracted from the accident description text, hidden danger entities related to the accident entities are obtained, and fault tree analysis is performed on the accident entities and hidden danger entities to obtain the accident cause analysis fault tree corresponding to the accident.
[0006] Preferably, a word vector algorithm is used to obtain the hidden danger entity corresponding to the accident entity within a set similarity threshold range.
[0007] As an improvement, the text-based accident cause analysis method includes the following steps:
[0008] S1. Obtain the accident description text;
[0009] S2. Perform word segmentation on the accident description text to obtain each accident entity, and use each accident entity as the top event and some intermediate events in the fault tree;
[0010] S3, using a word vector algorithm to calculate the similarity between each accident entity and each hidden danger entity in the hidden danger knowledge graph, and then obtaining each related hidden danger entity whose similarity with each accident entity is within a threshold range, wherein the hidden danger knowledge graph contains association relationship data between various hidden danger entities and hidden danger texts;
[0011] S4. Taking each relevant hidden danger entity obtained as an intermediate event of the fault tree;
[0012] S5. For each accident entity and each related hidden danger entity, based on the first mapping relationship between the fault entity and each safety concept in the enterprise safety management system, search for the corresponding safety concept in the management system knowledge graph; based on the second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard, search for the corresponding safety concept in the industry best practice safety standard; the management system knowledge graph contains the corresponding relationship between the management system name / clause and the safety concept;
[0013] S6. For accident entities and hidden danger entities that can be searched for corresponding safety concepts in the management system knowledge graph, the clauses corresponding to the corresponding safety concepts in the enterprise safety management system are used as the bottom events of the fault tree;
[0014] For accident entities and hidden danger entities whose corresponding safety concepts can be found in the industry best practice safety standards, the corresponding safety concepts in the industry best practice safety standards are used as the bottom events of the fault tree;
[0015] For accident entities and hidden danger entities whose corresponding safety concepts cannot be found in the management system knowledge graph and industry best practice safety standards, S7 is performed;
[0016] S7. Using a safety dictionary containing mapping relationships between various fault entities and safety concepts, obtain the safety concepts corresponding to these accident entities and hidden danger entities, and search the currently constructed fault tree for these safety concepts to determine whether these safety concepts exist in the fault tree;
[0017] If yes, then build the relationship between these accident entities and hidden danger entities and the corresponding intermediate events and bottom events in the fault tree;
[0018] If not, the negative expressions of these safety concepts are added as the bottom events of the fault tree;
[0019] S8. Obtain a fault tree for analyzing the cause of the accident.
[0020] In order to improve the accuracy of the fault tree and make the analysis of the accident more accurate, in S8, after the basic fault tree is formed, the safety expert will sort out the basic fault tree, correct the erroneous content of the fault tree, form a correct fault tree, and complete the construction of the fault tree for accident cause analysis.
[0021] Preferably, the method for constructing the hidden danger knowledge graph is:
[0022] S10, obtaining hidden danger text data in the hidden danger database;
[0023] S20, performing word segmentation processing on the hidden danger text data, and then obtaining each hidden danger entity;
[0024] S30. Use a safety dictionary that contains comprehensive safety management content to construct a hidden danger knowledge graph that contains the corresponding relationship between each hidden danger entity and the hidden danger text.
[0025] Preferably, a security dictionary containing comprehensive security management content is used to construct a management system knowledge graph that contains the correspondence between management system names / clauses and security concepts.
[0026] Preferably, a first mapping relationship between a fault entity and each security concept in the enterprise security management system and a second mapping relationship between each fault entity and each security concept in the industry best practice security standard are pre-established and pre-stored in the security dictionary.
[0027] Compared with the prior art, the advantages of the present invention are: the text-based accident cause analysis method of the present invention, when constructing a fault tree analysis of the cause of an accident, not only performs a fault tree analysis on the accident entity corresponding to the accident itself, but also performs a fault tree analysis on the hidden danger subject related to the accident entity. The accident cause analysis fault tree obtained in this way has a more comprehensive and complete analysis of the cause of the accident. The fault cause query based on the accident cause analysis fault tree is not prone to missed detection. After the corresponding cause is rectified, the probability of similar safety accidents will be greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 The figure is a flowchart of a text-based accident cause analysis method in an embodiment of the present invention. DETAILED DESCRIPTION
[0029] The present invention is further described in detail below with reference to the accompanying drawings.
[0030] like Figure 1As shown, the core content of the text-based accident cause analysis method in this embodiment is: extracting accident entities based on the accident description text, obtaining hidden danger entities related to the accident entities, and performing fault tree analysis on the accident entities and hidden danger entities to obtain the accident cause analysis fault tree corresponding to the accident. The acquisition of relevant hidden danger entities can be calculated using various algorithms in the prior art. For example, the word vector algorithm in the prior art can be used to obtain the hidden danger entity corresponding to the accident entity within the set similarity threshold range. The similarity threshold range can be set according to the specific situation.
[0031] The concept of accident entity is the substantive keyword in the accident text description. For example, in the accident text description of "the staff's hands were scalded", the accident entity is "scald". The concept of hidden danger entity is the substantive keyword in the hidden danger text description. For example, in the hidden danger text description of "if the guardrail is not installed, the staff will fall and get injured", the hidden danger entity is "fall".
[0032] Before implementing the text-based accident cause analysis method in this embodiment, a security dictionary will be pre-stored, and the security dictionary can be set according to specific needs, such as including all security management content that can be currently obtained in the field of security management. It can also be a security dictionary that includes comprehensive security management content in a specific technical field of the user applying the text-based accident cause analysis method.
[0033] The safety dictionary includes mapping relationships between various fault entities and safety concepts, while accident entities and hidden danger entities are included in the fault entity database.
[0034] In addition, for the users of this method, the first mapping relationship between each fault entity and each safety concept in the enterprise safety management system and the second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard are additionally established, and the first mapping relationship between each fault entity and each safety concept in the enterprise safety management system and the second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard are placed in the security dictionary database. The contents of the accident entity and the hidden danger entity are both included in the fault entity of the security dictionary.
[0035] The text-based accident cause analysis method in this embodiment specifically includes the following steps.
[0036] S1. Obtain the accident description text.
[0037] S2. Perform word segmentation on the accident description text to obtain each accident entity, and use each accident entity as the top event and some intermediate events in the fault tree. A paragraph of accident description text may contain multiple accident entities, and whether each accident entity is used as a fixed time or an intermediate event of the fault tree is determined by setting a corresponding algorithm according to the specific application environment, thereby determining whether each accident entity is used as a fixed time or an intermediate event of the fault tree. The word segmentation processing of the accident description text can adopt various word segmentation processing algorithms in the prior art.
[0038] S3. Use the word vector algorithm to calculate the similarity between each accident entity and each hidden danger entity in the hidden danger knowledge graph, and then obtain each related hidden danger entity whose similarity with each accident entity is within a threshold range. The hidden danger knowledge graph contains the association relationship data between various hidden danger entities and hidden danger texts.
[0039] The method for constructing the hidden danger knowledge graph in this embodiment is:
[0040] S10, obtaining hidden danger text data in a hidden danger database; the hidden danger database can be constructed by using the method and applying various files related to safety hidden dangers currently stored in the company;
[0041] S20, performing word segmentation processing on the hidden danger text data, and then obtaining each hidden danger entity; the word segmentation processing of the hidden danger text data can adopt various word segmentation processing algorithms in the prior art.
[0042] S30. Use a safety dictionary that contains comprehensive safety management content to construct a hidden danger knowledge graph that contains the corresponding relationship between each hidden danger entity and the hidden danger text.
[0043] S4. Take the obtained relevant hidden danger entities as the intermediate events of the fault tree.
[0044] S5. For each accident entity and each related hidden danger entity, based on the first mapping relationship between the fault entity and each safety concept in the enterprise safety management system, search for the corresponding safety concept in the management system knowledge graph; based on the second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard, search for the corresponding safety concept in the industry best practice safety standard. Using the safety dictionary containing comprehensive safety management content, construct a management system knowledge graph that contains the corresponding relationship between the management system name / clause and the safety concept. The management system knowledge graph contains the corresponding relationship between the management system name / clause and the safety concept.
[0045] S6. For accident entities and hidden danger entities whose corresponding safety concepts can be searched in the management system knowledge graph, the clauses corresponding to the corresponding safety concepts in the enterprise safety management system will be used as the bottom events of the fault tree.
[0046] For accident entities and hidden danger entities whose corresponding safety concepts can be found in the industry best practice safety standards, the corresponding safety concepts in the industry best practice safety standards are used as the bottom events of the fault tree;
[0047] For accident entities and hidden danger entities whose corresponding safety concepts cannot be found in the management system knowledge graph and industry best practice safety standards, S7 is performed.
[0048] S7. Utilize the safety dictionary that contains the mapping relationship between various fault entities and safety concepts to obtain the safety concepts corresponding to these accident entities and hidden danger entities, and search the currently constructed fault tree for these safety concepts to determine whether these safety concepts exist in the fault tree.
[0049] If so, then build the relationship between these accident entities and hidden danger entities and the corresponding intermediate events and bottom events in the fault tree.
[0050] If not, the negation of these safety concepts is added as the bottom event of the fault tree.
[0051] S8. Obtain the fault tree for analyzing the cause of the accident. In order to improve the accuracy of the fault tree and thus make the analysis of the accident more accurate, a basic fault tree can be formed first, and then the safety expert can sort out the basic fault tree and correct the erroneous content of the fault tree. The correction work includes proposing or correcting work to form a correct fault tree and complete the construction of the fault tree for analyzing the cause of the accident.
[0052] The text-based accident cause analysis method in this embodiment, when constructing a fault tree analysis of the cause of an accident, not only performs a fault tree analysis on the accident entity corresponding to the accident itself, but also performs a fault tree analysis on the hidden danger subject related to the accident entity. The accident cause analysis fault tree obtained in this way has a more comprehensive and complete analysis of the cause of the accident. The fault cause query based on the accident cause analysis fault tree is not prone to missed detection. After the corresponding cause is rectified, the probability of similar safety accidents will be greatly reduced.
Claims
1. A text-based accident cause analysis method, characterized in that: The accident entity is extracted from the accident description text, the hidden danger entity related to the accident entity is obtained, and the fault tree analysis is performed on the accident entity and the hidden danger entity to obtain the accident cause analysis fault tree corresponding to the accident; specifically, the word vector algorithm is used to obtain the hidden danger entity corresponding to the accident entity within the set similarity threshold range; The text-based accident cause analysis method comprises the following steps: S1. Obtain the accident description text; S2. Perform word segmentation on the accident description text to obtain each accident entity, and use each accident entity as the top event and some intermediate events in the fault tree; S3, using a word vector algorithm to calculate the similarity between each accident entity and each hidden danger entity in the hidden danger knowledge graph, and then obtaining each related hidden danger entity whose similarity with each accident entity is within a threshold range, wherein the hidden danger knowledge graph contains association relationship data between various hidden danger entities and hidden danger texts; S4. Taking each relevant hidden danger entity obtained as an intermediate event of the fault tree; S5. For each accident entity and each related hidden danger entity, based on the first mapping relationship between the fault entity and each safety concept in the enterprise safety management system, search for the corresponding safety concept in the management system knowledge graph; based on the second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard, search for the corresponding safety concept in the industry best practice safety standard; the management system knowledge graph contains the corresponding relationship between the management system name / clause and the safety concept; S6. For accident entities and hidden danger entities that can be searched for corresponding safety concepts in the management system knowledge graph, the clauses corresponding to the corresponding safety concepts in the enterprise safety management system are used as the bottom events of the fault tree; For accident entities and hidden danger entities whose corresponding safety concepts can be found in the industry best practice safety standards, the corresponding safety concepts in the industry best practice safety standards are used as the bottom events of the fault tree; For accident entities and hidden danger entities whose corresponding safety concepts cannot be found in the management system knowledge graph and industry best practice safety standards, S7 is performed; S7. Using a safety dictionary containing mapping relationships between various fault entities and safety concepts, obtain the safety concepts corresponding to these accident entities and hidden danger entities, and search the currently constructed fault tree for these safety concepts to determine whether these safety concepts exist in the fault tree; If yes, then build the relationship between these accident entities and hidden danger entities and the corresponding intermediate events and bottom events in the fault tree; If not, the negative expressions of these safety concepts are added as the bottom events of the fault tree; S8. Obtain a fault tree for analyzing the cause of the accident.
2. The text-based accident cause analysis method according to claim 1, characterized in that: In S8, after forming the basic fault tree, the safety expert will sort out the basic fault tree, correct the erroneous content of the fault tree, form a correct fault tree, and complete the construction of the fault tree for accident cause analysis.
3. The text-based accident cause analysis method according to claim 1 or 2, characterized in that: The method for constructing the hidden danger knowledge graph is as follows: S10, obtaining hidden danger text data in the hidden danger database; S20, performing word segmentation processing on the hidden danger text data, and then obtaining each hidden danger entity; S30. Use a safety dictionary that contains comprehensive safety management content to construct a hidden danger knowledge graph that contains the corresponding relationship between each hidden danger entity and the hidden danger text.
4. The text-based accident cause analysis method according to claim 1 or 2, characterized in that: Utilize a security dictionary that contains comprehensive security management content to construct a management system knowledge graph that shows the correspondence between management system names / clauses and security concepts.
5. The text-based accident cause analysis method according to claim 1 or 2, characterized in that: A first mapping relationship between a fault entity and each safety concept in the enterprise safety management system and a second mapping relationship between each fault entity and each safety concept in the industry best practice safety standard are pre-established and pre-stored in the security dictionary.
