CN-122019531-A - Pressurized water reactor system fault tree building method based on large language model

CN122019531ACN 122019531 ACN122019531 ACN 122019531ACN-122019531-A

Abstract

A pressurized water reactor system fault tree building method based on a large language model relates to the field of fault tree analysis. The method comprises the steps of outputting structured module failure reason items according to system success criteria, a system hierarchy, an equipment list, connection relations and equipment failure modes of a pressurized water reactor system, setting overall failure of the system as a top event based on the module failure reason items, outputting subsystem failure events, continuously expanding and mapping the system to the module failure intermediate events which cause the subsystem failure based on the subsystem failure events, outputting the module failure events, tracing back to specific equipment and failure modes thereof one by one based on the module failure events, generating equipment failure bottom events, constructing a preliminary failure tree and outputting the preliminary failure tree, and generating a compact final failure tree.

Inventors

DING MING
HAO XIAOTIAN
YANG YONGYONG
CAO XIAXIN
GUO ZEHUA
MENG ZHAOMING
HE YU
WANG YANKAI

Assignees

哈尔滨工程大学

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (10)

1. A pressurized water reactor system fault tree building method based on a large language model is characterized by comprising the following steps: A step of analyzing and reasoning module failure reasons by using a large language model and outputting structured module failure reason items as follow-up input by taking a system success criterion, a system hierarchical structure, an equipment list, a connection relation and an equipment failure mode of the pressurized water reactor system as input information; Setting the overall failure of the system as a top event based on the module failure reason item, mapping the top event to a subsystem failure intermediate event which causes the system failure, and outputting the subsystem failure event as a subsequent input step; continuously expanding and mapping the subsystem failure event to a module failure intermediate event which causes subsystem failure based on the subsystem failure event, and outputting the module failure event as a subsequent input; Based on the module failure event, tracing to specific equipment and failure modes thereof one by one, generating an equipment failure bottom event, constructing a preliminary fault tree and outputting the preliminary fault tree as a subsequent input; And (3) performing a cleaning operation based on the preliminary fault tree, and automatically eliminating repeated events, equivalent branches and redundant paths to generate a compact final fault tree.
2. The method for building a pressurized water reactor system fault tree based on a large language model according to claim 1, wherein the system information comprises system success criteria, a system hierarchy, a device list, a connection relationship, and a device failure mode.
3. The method for building the pressurized water reactor system fault tree based on the large language model according to claim 1, wherein the top event is a system overall failure event, when the top event is mapped to a subsystem failure intermediate event, if a specific failure reason is searched, an entry is directly generated, and if the specific failure reason is not searched, the subsystem is uniformly marked as failed.
4. The method for building the pressurized water reactor system fault tree based on the large language model according to claim 1, wherein when the subsystem failure event is unfolded downwards to be a module failure intermediate event, if the specific failure reason is searched, an entry is directly generated, and if the specific failure reason is not searched, the module is uniformly marked as failed.
5. The pressurized water reactor system fault tree building method based on the large language model according to claim 1, wherein when the module failure event is traced back to a specific equipment failure bottom event one by one, the equipment failure and the failure form thereof are generated by combining the module failure reason item output by the large language model.
6. The method for building the pressurized water reactor system fault tree based on the large language model according to claim 1, wherein the final fault tree is compact in structure and clear in logic, and is suitable for reliability analysis and probability safety evaluation of the pressurized water reactor system.
7. The utility model provides a pressurized water reactor system fault tree building device based on big language model which characterized in that includes: The system success rule, the system hierarchy, the equipment list, the connection relation and the equipment failure mode of the pressurized water reactor system are taken as input information, the module failure reasons are analyzed and inferred by the large language model, and the structured module failure reason items are output as modules for subsequent input; setting the overall failure of the system as a top event based on the module failure reason item, mapping the top event to a subsystem failure intermediate event which causes the system failure, and outputting the subsystem failure event as a module of a subsequent input; Based on the subsystem failure event, continuing to expand and map to a module failure intermediate event which causes subsystem failure, and outputting the module failure event as a module for subsequent input; based on the module failure event, tracing to specific equipment and failure modes thereof one by one, generating an equipment failure bottom event, constructing a preliminary fault tree and outputting the preliminary fault tree as a module for subsequent input; Based on the preliminary fault tree, a cleaning operation is performed to automatically eliminate duplicate events, equivalent branches and redundant paths, resulting in a compact module of the final fault tree.
8. Computer storage medium for storing a computer program, characterized in that the computer performs the method of claim 1 when the computer program is read by the computer.
9. A computer comprising a processor and a storage medium, characterized in that the computer performs the method of claim 1 when the processor reads a computer program stored in the storage medium.
10. Computer program product, as a computer program, characterized in that the method of claim 1 is implemented when the computer program is executed.

Description

Pressurized water reactor system fault tree building method based on large language model Technical Field Relates to the field of fault tree analysis, in particular to a pressurized water reactor system fault tree building based on a large language model. Background Fault tree analysis, which is a typical system reliability analysis method, has been widely used in high security industries such as nuclear energy, aerospace, chemical industry, aviation, etc., since the proposal of the 60 th century. The basic idea is to decompose the top-level failure event of the system into bottom events of the subsystem, module and equipment level step by step through logical reasoning, thereby revealing the mechanism and path of the system failure. In the nuclear power field, particularly in a pressurized water reactor system, fault tree analysis is one of important basic tools of Probability Safety Analysis (PSA) and can provide decision basis for system design, operation maintenance and risk assessment. The existing fault tree construction method mainly depends on a manual mode. Researchers or engineering specialists typically develop fault trees based on an understanding of the system architecture and failure mechanisms, combined with empirical deductions. The method is visual, but depends on expert knowledge seriously, the tree building process is long in time consumption and low in efficiency, and the result is easily influenced by subjective factors. For example, in analysis of pressurized water reactor cooling systems, engineers need to manually identify failure modes of each device and their impact on system functions, and problems of omission or logic inconsistency are very likely to occur when the system is large in scale and the failure path is complex. With the development of artificial intelligence, students have tried to assist in fault tree construction by using knowledge-graph, machine learning and other methods. For example, there are studies on an automatic tree building method based on a knowledge base, which is to semi-automatically generate a fault tree structure by establishing a failure mode database in advance, and a literature exploration method which uses a natural language processing method based on deep learning to extract and map failure information in a device description document into a fault tree model. However, these methods have certain limitations in the face of highly complex, multiple coupling effects of the engineering objects such as pressurized water reactor systems. On one hand, the traditional knowledge graph method is difficult to cover all potential equipment interaction relations and is easy to miss collaborative fault paths, and on the other hand, the deep learning method can automatically extract partial failure information, but has insufficient capabilities in the aspects of logic reasoning and hierarchical modeling, so that the generated fault tree has redundant branches or does not accord with the engineering actual problem. In summary, the fault tree construction process in the prior art has the defects of dependence on manpower, low efficiency and insufficient logic consistency, and is difficult to cope with the multi-level failure path and multiple fault synergistic effect in the complex systems such as pressurized water reactor and the like. Disclosure of Invention In order to solve the defects that the fault tree construction process in the prior art depends on manpower, has low efficiency and insufficient logic consistency, and is difficult to cope with the synergistic effect of multi-level failure paths and multiple faults in complex systems such as pressurized water reactors, the invention provides the following technical scheme: A pressurized water reactor system fault tree building method based on a large language model comprises the following steps: A step of analyzing and reasoning module failure reasons by using a large language model and outputting structured module failure reason items as follow-up input by taking a system success criterion, a system hierarchical structure, an equipment list, a connection relation and an equipment failure mode of the pressurized water reactor system as input information; Setting the overall failure of the system as a top event based on the module failure reason item, mapping the top event to a subsystem failure intermediate event which causes the system failure, and outputting the subsystem failure event as a subsequent input step; continuously expanding and mapping the subsystem failure event to a module failure intermediate event which causes subsystem failure based on the subsystem failure event, and outputting the module failure event as a subsequent input; Based on the module failure event, tracing to specific equipment and failure modes thereof one by one, generating an equipment failure bottom event, constructing a preliminary fault tree and outputting the preliminary fault tree as a subsequent input; And (3) performi