CN-122024546-A - Lifting crane fault maintenance training system based on large language model

CN122024546ACN 122024546 ACN122024546 ACN 122024546ACN-122024546-A

Abstract

The invention discloses a crane fault maintenance training system based on a large language model, and belongs to the field of industrial equipment maintenance training. The method aims to solve the problems that the existing training mode is large in potential safety hazard, single in fault scene, dependent on experience in diagnosis and lacking in closed loop assessment. The system comprises a PLC communication and state synchronization module, a fault injection and simulation module, a UE5 three-dimensional simulation and variable mapping module, an intelligent retrieval and LLM diagnosis module, an assessment and evaluation module and an automatic assessment and quantization scoring module, wherein the PLC communication and state synchronization module is used for interacting with a PLC in real time, the fault injection and simulation module is used for simulating various faults according to rules, the UE5 three-dimensional simulation and variable mapping module is used for constructing a three-dimensional environment and realizing virtual and real synchronization, the intelligent retrieval and LLM diagnosis module is used for providing fault diagnosis and maintenance guidance based on a local knowledge base and real-time data. The intelligent training system has the beneficial effects that high-fidelity and randomly configurable fault simulation is realized, safe and efficient offline diagnosis guidance is provided through the localization AI, and the safety, the authenticity and the effect scalability of training are remarkably improved by utilizing a three-dimensional immersive environment and a closed-loop evaluation system.

Inventors

PENG CHEN
WANG LIJUN
WU WENBIN
WEI DONGPEI
LIU KAIDONG
ZOU BING
GAO PING
ZHANG DONGXU

Assignees

建龙(辽宁)节能环保科技有限公司
抚顺新钢铁有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260122

Claims (8)

1. The crane fault maintenance training system based on the large language model is characterized by comprising a PLC communication and state synchronization module, a fault injection and simulation module, a UE5 three-dimensional simulation and variable mapping module, an intelligent retrieval and LLM diagnosis module and an assessment and evaluation module; The PLC communication and state synchronization module is used for establishing real-time communication connection with a Programmable Logic Controller (PLC) of the crane, reading and writing I/O area and data block data, writing fault zone bits, reading DB block data and synchronizing equipment states; The fault injection and simulation module is used for configuring a fault configuration tool at the PLC end of the crane, and writing fault data into the PLC communication and state synchronization module according to a user-defined or randomly generated fault rule so as to simulate the fault type of the crane in a virtual environment; The UE5 three-dimensional simulation and variable mapping module is used for constructing a three-dimensional virtual simulation environment of the crane, mapping the real-time data acquired by the PLC communication and state synchronization module to the state and animation of the virtual equipment, and realizing virtual-real synchronization of the PLC state and the animation of the virtual equipment; The intelligent retrieval and LLM diagnosis module is used for carrying out intention recognition and reasoning on fault description input by a user by combining real-time operation data acquired by the PLC communication and state synchronization module based on a structured knowledge base and a retrieval enhancement generation method which are constructed locally, and generating guide information comprising fault diagnosis and investigation steps; The assessment and evaluation module automatically generates assessment questions and standard answers according to the current fault scene, records and analyzes operation logs of students in the UE5 three-dimensional simulation and variable mapping module, quantitatively scores the operation logs from multiple dimensions and generates an evaluation report.
2. The system of claim 1, wherein the PLC communication and status synchronization module specifically comprises: the communication configuration unit is used for establishing communication protocol connection with the PLC and carrying out parameter configuration; and the data interaction optimizing unit adopts a read-write request batch processing, priority-based queue scheduling, overtime retry and disconnection automatic reconnection mechanism to ensure the real-time performance and reliability of communication.
3. The system of claim 1, wherein the fault injection and simulation module specifically comprises: The fault configuration unit provides a graphical interface for a user to define the address range, the type, the occurrence probability, the duration and the association relation among faults; The random fault generation unit adopts a poisson process or a non-homogeneous probability model to simulate the burstiness and dynamic probability of faults and triggers the faults based on a predefined fault type library and association rules; And the fault propagation control unit is used for determining whether to trigger the secondary fault according to the preset association probability after the primary fault is triggered, and recording a fault propagation chain.
4. The system of claim 1, wherein the intelligent retrieval and LLM diagnostic module specifically comprises: the local knowledge base construction unit is used for cleaning, segmenting and vectorizing the maintenance manual and the maintenance plan document to construct a vector database supporting semantic retrieval; the retrieval enhancement generation unit fuses keyword matching, semantic similarity and real-time operation data characteristics from the PLC, and dynamically rearranges retrieval results; The large language model reasoning unit is used for reasoning the query which fuses the search result and the real-time data based on the large language model, and generating structured diagnosis and guidance output; The diagnosis result optimizing unit is used for calculating the confidence coefficient of the diagnosis result and associating evidence sources, and comparing and forcedly checking the confidence coefficient and the associated evidence sources with a predefined safe operation step template.
5. The system of claim 1, wherein the UE5 three-dimensional simulation and variable mapping module further integrates an augmented reality/virtual reality AR/VR interface and a virtual character model, and the virtual character model performs voice broadcasting and operation demonstration through offline text-to-voice TTS and action driving according to the structured result output by the intelligent retrieval and LLM diagnostic module.
6. The system of claim 1, wherein the assessment questions generated by the assessment and evaluation module include real operation questions and theoretical questions, and the evaluation dimensions include operation step completion, critical safety step coverage, number of misoperation, task completion time and fault clearance rate.
7. A computer program product comprising a computer program which when executed by a processor implements the functionality of the large language model based lifting ride fault maintenance training system of any one of claims 1 to 6.
8. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the functionality of the large language model based crane fault maintenance training system of any of claims 1 to 6.

Description

Lifting crane fault maintenance training system based on large language model Technical Field The invention relates to the technical field of industrial equipment maintenance training, in particular to a crane fault maintenance training system based on a large language model. Background At present, aiming at the training and fault diagnosis of crane cranes and electrical equipment thereof, the teaching mode of combining traditional offline real operation with static data is mainly adopted, and the problems of large potential operational safety hazard, insufficient fault scene reduction degree, lower skill transmission efficiency and the like exist, so that the method is difficult to adapt to the urgent requirements of modern industry on the safe operation and efficient maintenance of equipment. Specifically, the prior art has the following disadvantages: The training process is mostly dependent on paper teaching materials, fixed videos or virtual simulation systems of preset scenes, the randomness and dynamic interaction capability of faults are lacked, sudden characteristics and diversity of the faults in a real working environment are difficult to simulate, and the students lack enough strain capability in actual coping. The existing virtual training system has single preset fault type, weak random generation capability, poor real-time interactivity with I/O and data block modules in the PLC system, difficulty in truly reproducing complex fault scenes of multi-source coupling, and lack of complete closed loop from fault injection to state feedback. The mainstream PLC system is generally focused on offline debugging of programs and basic I/O state monitoring, and failure to effectively integrate fault simulation functions of external electrical elements (such as contactors and frequency converters) leads to lack of effective linkage between fault simulation and system response, so that realism of training and diagnosis is limited. When troubleshooting, maintenance personnel need to manually inquire and scatter the technical data (including equipment specifications, maintenance plans, historical cases and the like) at a plurality of places, the information acquisition path is long, and the diagnosis process is seriously dependent on personal experience due to non-uniform document version, content deletion or low retrieval efficiency, so that misjudgment is easy to occur. The current AI scheme applied to fault diagnosis mostly depends on a cloud large model, and cannot fully combine a localization Large Language Model (LLM) and an enterprise exclusive knowledge base (such as a maintenance manual, a maintenance standard and the like), and meanwhile, a three-dimensional visual simulation environment is not embedded, so that instant immersive guidance based on voice or text question cannot be realized. The lack of a systematic platform capable of uniformly scheduling key functions such as S7 communication, random fault simulation, LLM decision support, UE5 interaction and the like causes that links of training teaching, fault diagnosis and effect assessment are mutually disjointed, and a closed loop system covering the whole process of equipment management is difficult to construct. Disclosure of Invention Technical problem to be solved In view of the defects and shortcomings of the prior art, the invention provides a lifting crane fault maintenance training system based on a large language model, which solves the technical problems that in the lifting crane and electrical equipment training and fault diagnosis thereof, the traditional mode of combining offline real operation with static data has large potential safety hazard, insufficient fault scene reduction degree, low skill transmission efficiency and the like, and is difficult to adapt to the requirements of safe operation and efficient maintenance of modern industrial equipment. Technical proposal In order to achieve the above purpose, the main technical scheme adopted by the invention comprises the following steps: The invention provides a crane fault maintenance training system based on a large language model, which comprises a PLC communication and state synchronization module, a fault injection and simulation module, a UE5 three-dimensional simulation and variable mapping module, an intelligent retrieval and LLM diagnosis module and an assessment and evaluation module; The PLC communication and state synchronization module is used for establishing real-time communication connection with a Programmable Logic Controller (PLC) of the crane, reading and writing I/O area and data block data, writing fault zone bits, reading DB block data and synchronizing equipment states; The fault injection and simulation module is used for configuring a fault configuration tool at the PLC end of the crane, and writing fault data into the PLC communication and state synchronization module according to a user-defined or randomly generated fault rule so as to simulate the