CN-122019604-A - System information intelligent acquisition method based on collaboration of multi-mode large model and large language model

CN122019604ACN 122019604 ACN122019604 ACN 122019604ACN-122019604-A

Abstract

A system information intelligent acquisition method based on the cooperation of a multi-mode large model and a large language model relates to the field of fault tree analysis. The method comprises the steps of obtaining and preprocessing a system flow chart, generating high-quality input data through functional decoupling and physical boundary decomposition, carrying out symbol recognition and text extraction to obtain preliminary topology information, forming a standardized system information file through a regular expression, inputting the file into a large language model to obtain a fault information set containing a fault mode and tracing information, integrating the topology file and the fault information to form system information data, and outputting the standardized file which can be directly used for fault tree analysis and reliability assessment through consistency and integrity verification. The method is suitable for system topology construction, fault tree analysis and safety reliability assessment of a complex industrial system.

Inventors

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

Assignees

哈尔滨工程大学

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (10)

1. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model is characterized by comprising the following steps: Acquiring and preprocessing a system flow chart, decomposing the functional decoupling property and the physical boundary property of the system flow chart, and generating high-quality input data through a Base64 coding and resolution self-adaption technology; performing symbol recognition and text extraction on the input data by using a multi-mode large model to obtain preliminary topology information comprising equipment numbers, equipment names, parameters and connection sequences; performing regular expression extraction and JSON formatting on the preliminary topology information, removing redundant or conflict information, and outputting a standardized system information file; Inputting the standardized system information file into a large language model, carrying out vectorization retrieval and semantic matching on a system information knowledge base, and outputting a fault information set containing a fault mode and tracing information; Integrating the standardized system information file and the fault information set to form a system information data comprising equipment, connection and fault triad; And checking the uniqueness, the integrity and the logical consistency of the system information data, and outputting a standardized system information file which can be directly used for fault tree analysis and reliability evaluation.
2. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model according to claim 1 is characterized in that in the process of acquiring and preprocessing a system flow chart, the modularization processing of the flow chart is realized through the double decomposition of functional decoupling and physical boundary, symbol ambiguity is reduced, and recognition accuracy is improved.
3. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model according to claim 1, wherein character and flow double-track system prompt word paradigm constraint output logic is adopted in the steps of carrying out symbol recognition and text extraction by utilizing the multi-mode large model so as to ensure consistency and standardization of equipment numbers, equipment names and connection relations.
4. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model according to claim 1, wherein in the process of carrying out regular expression extraction and JSON formatting on the preliminary topology information, uniqueness verification, integrity verification and logic consistency verification are carried out so as to ensure the reliability of the standardized system information file.
5. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model according to claim 1 is characterized in that in the process of vectorizing retrieval and semantic matching of a system information knowledge base, a deep neural network embedded model is adopted to convert texts into high-dimensional semantic vectors, and semantic matching is completed through a similarity algorithm.
6. The intelligent system information acquisition method based on the collaboration of the multi-mode large model and the large language model according to claim 1, wherein in the process of integrating the standardized system information file and the fault information set, a unified representation of equipment-connection-fault is formed by establishing a mapping relation between equipment and a fault mode.
7. The intelligent system information acquisition device based on the cooperation of the multi-mode large model and the large language model is characterized by comprising: A module for acquiring and preprocessing a system flow chart, decomposing the functional decoupling property and the physical boundary property of the system flow chart, and generating high-quality input data through a Base64 coding and resolution self-adaption technology; Performing symbol recognition and text extraction on the input data by using a multi-mode large model to obtain a module containing preliminary topology information of equipment numbers, equipment names, parameters and connection sequences; The module is used for extracting regular expressions and formatting JSON (java server on) on the preliminary topology information, eliminating redundant or conflict information and outputting a standardized system information file; The standardized system information file is input into a large language model, vectorization retrieval and semantic matching are carried out on a system information knowledge base, and a fault information set module containing fault modes and tracing information is output; Integrating the standardized system information file and the fault information set to form a module comprising equipment, connection and fault trinity system information materials; And (3) checking the uniqueness, the integrity and the logical consistency of the system information data, and outputting a module of a standardized system information file which can be directly used for fault tree analysis and reliability evaluation.
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

System information intelligent acquisition method based on collaboration of multi-mode large model and large language model Technical Field The method relates to the field of fault tree analysis, in particular to intelligent acquisition of system information based on collaboration of a multi-mode large model and a large language model. Background In the design, operation and maintenance processes of a complex industrial system, obtaining system structure information and equipment fault knowledge is a precondition for fault tree analysis and safety evaluation. Currently, there have been several studies in the art attempting to automatically acquire and process system information. On the one hand, research based on image recognition is more common, for example, character and symbol in engineering drawings or flowcharts are recognized by utilizing OCR technology, and connection relation between devices is extracted by a computer vision method. The method improves the information processing efficiency to a certain extent, but is often limited by drawing symbol complexity, text ambiguity and multi-source image format difference, has limited recognition precision, and is difficult to ensure the integrity of structural information. On the other hand, research based on text mining and natural language processing is also increasingly applied to system information processing. For example, there are research attempts to extract device names, parameters, and failure modes from design specifications, operation and maintenance manuals, or incident reports through keyword matching, entity recognition, and rule parsing. Such methods can support knowledge acquisition to some extent, but have the disadvantage of lacking deep semantic understanding, difficulty in correctly handling ambiguity of terms of art, and inability to establish topological associations between devices. In recent years, part of the work explores cross-modal information processing methods, such as joint training of images and text features, to achieve multi-modal information extraction. However, these methods still have the problems of insufficient generalization and poor suitability for complex industrial systems, and often require a lot of manual intervention for post-processing in engineering applications, which makes it difficult to achieve efficient and standardized information acquisition. In summary, the prior art has the defects of low efficiency, insufficient accuracy and lack of cross-mode co-processing capability in acquiring the system topology structure and fault knowledge. Disclosure of Invention In order to solve the defects of low efficiency, insufficient accuracy and lack of cross-mode cooperative processing capability in the prior art for acquiring the topological structure and fault knowledge of a system, the invention provides the following technical scheme: A system information intelligent acquisition method based on the cooperation of a multi-mode large model and a large language model comprises the following steps: Acquiring and preprocessing a system flow chart, decomposing the functional decoupling property and the physical boundary property of the system flow chart, and generating high-quality input data through a Base64 coding and resolution self-adaption technology; performing symbol recognition and text extraction on the input data by using a multi-mode large model to obtain preliminary topology information comprising equipment numbers, equipment names, parameters and connection sequences; performing regular expression extraction and JSON formatting on the preliminary topology information, removing redundant or conflict information, and outputting a standardized system information file; Inputting the standardized system information file into a large language model, carrying out vectorization retrieval and semantic matching on a system information knowledge base, and outputting a fault information set containing a fault mode and tracing information; Integrating the standardized system information file and the fault information set to form a system information data comprising equipment, connection and fault triad; And checking the uniqueness, the integrity and the logical consistency of the system information data, and outputting a standardized system information file which can be directly used for fault tree analysis and reliability evaluation. Further, a preferred embodiment is provided, in the process of acquiring and preprocessing the flow chart of the system, the modularization processing of the flow chart is realized through the double decomposition of the functional decoupling property and the physical boundary property, so that the symbol ambiguity is reduced, and the recognition precision is improved. Further, a preferred embodiment is provided, in the steps of performing symbol recognition and text extraction by using the multi-mode large model, character and flow double-track system prompt word paradigm constraint ou