A digital training method and system based on safety production

Abstract

The application belongs to the technical field of electric power safety production, and discloses a digital training method and system based on safety production, wherein a micro-macro dual-domain mapping is constructed through a quantum-enhanced digital twin unit, so that students can directly observe the micro-processes such as the molecular chain fracture of the cable insulation layer and the evolution of the control cabinet arc discharge plasma, and the micro-processes are associated to macro fault performance; the problem that the traditional system can only present the surface state of the equipment is solved, the predication ability of the hidden danger of the high-voltage equipment is improved from the root, and the operation omission caused by the insufficient risk cognition is avoided; through a brain-computer collaborative cognitive training unit, the system can monitor the neural activity and operation characteristics of the students in real time, the cognitive load is precisely regulated by adjusting the stress parameters of the virtual scene; the decision delay problem of the students in the real high-voltage operation can be effectively relieved, the training effect is closer to the actual working condition, and the stable judgment and execution ability of the students when facing the high-risk operation is ensured.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric power safety production, and specifically relates to a digital training method and system based on safety production. BACKGROUND

[0002] In the field of intelligent power control equipment and cable manufacturing, safety production training is directly related to the operation standardization of high-voltage equipment, fault emergency handling capability and production efficiency. This field covers key links such as 35kV and above cable laying, intelligent control cabinet debugging, and insulation material performance testing. The operation process involves risk factors such as high voltage and high tension, and requires high skill proficiency and risk prediction ability of practitioners.

[0003] After searching, the invention patent with the publication number CN119963375A discloses a safety production digital training management system. Although it realizes personalized training scheme pushing by analyzing the behavior data such as the learning duration and examination results of students, it still has significant technical limitations: first, the system can only evaluate the student's state based on macro behavior data, and does not integrate physiological signals such as electroencephalogram (EEG) and electromyogram to build a cognitive-action correlation model. It cannot capture the micro cognitive state of students in high-voltage operation such as attention fluctuation and emotional tension in real time, resulting in insufficient matching of psychological state between virtual training and real high-voltage environment, and decision delay of students when facing high-risk scenes such as 35kV cable pressure test; second, the digital twin module only simulates macro operating parameters of equipment (such as temperature and pressure), and does not involve visualization of micro mechanisms such as cable insulation layer molecular chain rupture and control cabinet arc discharge plasma evolution. The student's understanding of potential risks stays at the phenomenon level and it is difficult to understand the fundamental mechanism of fault occurrence; third, the fault scene design relies on a preset case library and does not build a cross-device correlation graph through a graph neural network. It cannot reproduce chain reactions such as "poor cable joint sealing → local discharge → insulation breakdown → control cabinet override trip", which is disconnected from the fault conduction mechanism of equipment interconnection in actual production. SUMMARY

[0004] The purpose of the present application is to provide a digital training method and system based on safety production to solve the problems raised in the background art.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical solution: a digital training system based on safety production, which comprises:

[0006] Quantum-enhanced digital twin unit: Based on the physical properties of the equipment, a multi-scale digital twin body is constructed, and the micro changes of the cable insulation layer are tracked through quantum point markers, and the macro behavior of the control cabinet is associated with the micro mechanism; Its high-speed interface synchronizes real-time parameters to the cross-domain fault chain generation unit, supports dynamic scene construction, while receiving updated data from the self-evolving knowledge base, continuously optimizes model accuracy, and provides physical data support for the system;

[0007] Brain-computer collaborative cognitive training unit: Integrates EEG and electromyography devices to capture students' neural activity and operation characteristics, builds a "cognition-action" model, detects cognitive abnormalities, sends instructions to the holographic operation memory unit to trigger scene adaptation, encrypts cognitive data to the self-evolving knowledge base unit, and also receives scene parameters from the cross-domain fault chain unit to dynamically adjust monitoring thresholds, achieving collaborative training of cognition and operation;

[0008] Cross-domain fault chain generation unit: Based on real-time parameters and historical cases from the self-evolving knowledge base unit, a device correlation graph is constructed using graph neural networks, and a fault chain scene is generated based on student operation data and virtual device status, which is synchronized to the holographic operation memory unit for rendering, and the rules are output to the self-evolving knowledge base unit for iteration;

[0009] Holographic operation memory unit: Records expert operations and generates holographic templates through light field imaging, and after receiving scene parameters from the cross-domain fault chain generation unit, it enables students to "operate on the same screen" as virtual experts; AR devices compare operation deviations, and exoskeleton devices provide tactile guidance, with intensity and frequency adjusted according to the student's state from the brain-computer collaborative cognitive training unit, and deviation data uploaded to the self-evolving knowledge base unit;

[0010] Self-evolving knowledge base unit: Integrates multi-source information including fault chains, holographic operations, and brain-computer collaboration, refines rules and models through federated learning, and updates them to the quantum-enhanced digital twin unit to optimize physical models, provides the latest rules for the cross-domain fault chain generation unit, and triggers expert template iteration.

[0011] Preferably, the quantum-enhanced digital twin unit is used for:

[0012] (1) Multi-scale modeling and micro mechanism restoration: Based on the physical properties of intelligent power equipment, a multi-scale digital twin body is constructed, and quantum point marker algorithm is used for cable insulation layer modeling to track virtual quantum point trajectories and dynamically present micro changes such as molecular chain rupture and crystal zone ablation under temperature and stress; Intelligent control cabinet associates macro behaviors such as circuit breaker action with micro mechanisms such as electron transition through electromagnetic simulation module, realizing multi-dimensional restoration of equipment state and providing physical basis for subsequent scene construction;

[0013] Quantum point diffusion coefficient formula:

[0014] ,

[0015] wherein: D is the quantum dot diffusion coefficient, with the unit of is the core index describing the virtual molecular motion activity of the cable insulation layer; in the digital twin scene, its value directly corresponds to the looseness of the microstructure of the insulation material, and the larger the value, the higher the risk of molecular chain rupture;

[0016] D0 represents the initial diffusion coefficient, and the reference value is determined by the type of cable insulation material, such as for cross-linked polyethylene cable, and for rubber insulated cable, which is obtained by material laboratory measurement and pre-loaded into the system;

[0017] represents the diffusion activation energy, with the unit of eV, representing the energy threshold required for molecular chain rupture of the insulation material; for 10kV cable, its value is usually between 0.3~0.5eV, and is dynamically corrected according to the aging degree of the material (15%~20% lower for aged cable than for new cable);

[0018] represents the Boltzmann constant , which is used to convert the temperature parameter into energy unit to ensure the dimensional consistency of the formula;

[0019] represents the virtual environment temperature, with the unit of K, which is synchronized with the real-time temperature of the cable production line (range 293~353K), and is collected from the temperature control module of the extruder through the device data interaction unit;

[0020] represents the mechanical stress, with the unit of MPa, corresponding to the tension during cable laying, such as 8~12MPa for overhead cable and 5~8MPa for underground direct-buried cable, which is generated by converting the laying angle (such as increasing the tension by 20% when ≥45°) and traction force parameters in the virtual scene;

[0021] represents the stress sensitivity coefficient, which is dimensionless and calibrated by 1000 times of cable stretching experiment (polyethylene material takes , reflecting the acceleration effect of mechanical force on molecular diffusion;

[0022] Effect: Quantify the coupling effect of temperature and mechanical stress on micro-movement, and the calculation result is used to drive the insulation layer treeing growth animation in the digital twin unit (triggering partial discharge simulation when ), and the real-time data is pushed to the cross-domain fault chain generation unit as the starting condition of the "insulation failure→control cabinet tripping" fault chain;

[0023] (2) Data linkage and model iteration optimization: The unit is built-in with a high-speed interface, which synchronizes real-time parameters such as cable insulation resistance and control cabinet temperature field to the cross-domain fault chain generation unit, supporting dynamic fault scene construction; at the same time, it receives device characteristic update data from the self-evolution knowledge base unit, continuously optimizes the physical precision of the twin model, and ensures that the virtual scene is highly consistent with the actual device characteristics.

[0024] Preferably, the brain-computer collaborative cognitive training unit is used for:

[0025] (1) Multi-dimensional state monitoring and model construction: integrate EEG acquisition and electromyographic sensing devices to capture real-time neural activity and operation characteristics of students; EEG module identifies cognitive states such as attention and emotion through frontal lobe electrodes; electromyographic devices judge hand operation stability, and core algorithms fuse data to construct a "cognition-action" correlation model to accurately identify cognitive load abnormalities in high-voltage operation and provide a basis for scene regulation;

[0026] Cognitive load index formula:

[0027] ,

[0028] In the formula: Cognitive load index, value range 0~1, 0.7 is the critical value, in high-voltage cable wiring training, Indicates that the student has a risk of operation failure, such as accidentally touching the live terminal when wiring 35kV cable, or mistakenly setting the parameter threshold when debugging the control cabinet;

[0029] represents the amplitude ratio of theta wave (4~8Hz) and beta wave (13~30Hz), theta wave enhancement reflects anxiety state, such as when facing 35kV equipment, beta wave enhancement reflects attention concentration, and the frontal lobe electrode array of the EEG headset is used to collect (sampling rate ;

[0030] represents the real-time electromyographic signal amplitude, unit μV, collected by wrist electromyographic sensor, reflecting the degree of hand operation tremor (normal operation , nervous ;

[0031] represents the maximum electromyographic signal amplitude, unit μV, obtained by initial test of students, such as electromyographic value when holding cable stripping pliers with full force, as a benchmark for individual operation limit;

[0032] represents the weight coefficient , calibrated based on the data of 100 electrician training, the influence of neural signals on operation safety is more significant, so Higher weight;

[0033] Function: To assess the physical and mental state of trainees in real time. When the intensity exceeds 0.7, the system sends a command to the holographic operation memory unit to reduce the audio-visual intensity of the high-voltage arc in the virtual scene (by 30%), and triggers holographic prompts for expert operation. The time-series data is encrypted and uploaded to a self-evolving knowledge base to optimize the difficulty curve of personalized training programs;

[0034] (2) Data interaction and dynamic adjustment and adaptation: When a cognitive abnormality is detected, an adjustment command is generated and sent to the holographic operation memory unit to trigger scenario adaptation. At the same time, the encrypted student cognitive data is transmitted to the self-evolving knowledge base unit to support personalized solution optimization; the scenario complexity parameters are received from the cross-domain fault chain generation unit to dynamically adjust the cognitive monitoring threshold and achieve accurate matching between the training scenario and the student's state.

[0035] Preferably, the cross-domain fault chain generation unit is used for:

[0036] (1) Construction of the association graph and scenario triggering: Based on the physical parameters of the quantum-enhanced digital twin unit and the historical cases of the self-evolving knowledge base, the device association graph is constructed through the graph neural network; the graph includes cable production parameters and control cabinet components, and the node weights are dynamically adjusted according to the linkage relationship. When training is started, the system triggers a chain failure scenario such as cable laying deviation based on the trainee's operation data and the virtual equipment status.

[0037] Fault association weight formula:

[0038] ,

[0039] In the formula: Represents device node To the node The fault association weight, with a value of 0 to 1, is used to quantify the fault propagation probability (e.g., show After the malfunction There is an 80% chance of triggering this.

[0040] Indicating historical data trigger The number of faults comes from the fault case library of the self-evolving knowledge base, such as the number of records for "abnormal pressure of cable extruder (i) → core deviation of insulation layer (j)";

[0041] Represents a node The physical parameters, For cable parameters, such as extruder pressure MPa, To control the cabinet parameters, such as the PLC temperature alarm threshold ;

[0042] represents the parameter reference value, the rated parameter for the normal operation of the equipment, such as the rated pressure of the extruder MPa, the rated temperature of the control cabinet , determined by the equipment manual;

[0043] Effect: Calculate the correlation strength of the equipment through the graph neural network, for example, when = 0.75, the system preferentially generates a fault chain of "extruder pressure exceeds the standard → cable resistance is abnormal → control cabinet over-current protection acts", the weight matrix is synchronized to the holographic operation memory unit, driving the linkage change of the equipment state in the virtual scene (such as the flashing of the control cabinet indicator light with the fluctuation of the cable parameter);

[0044] (2) Data synchronization and rule iteration: The generated dynamic fault chain data is synchronized to the holographic operation memory unit to drive scene rendering, and the reverse output fault evolution rules are output to the self-evolution knowledge base unit; By continuously receiving multi-source data, the correlation graph is optimized to ensure the coherence and timeliness of the fault scene, and to provide complex scenarios close to reality for training.

[0045] Preferably, the holographic operation memory unit is used for:

[0046] (1) Expert template construction and on-screen operation: Record expert operation data through light field imaging, form an interactive holographic template through three-dimensional reconstruction, receive dynamic fault chain data from the cross-domain fault chain generation unit during training, call the matching template, and realize "on-screen operation" of the student and the virtual expert through holographic projection; The AR device captures the student's actions in real time and compares them with the template to generate a deviation report;

[0047] (2) Haptic guidance and data feedback: The exoskeleton feedback device provides haptic guidance at key nodes such as cable crimping based on the comparison results, and the intensity and frequency are dynamically adjusted according to the student state of the brain-computer collaborative cognitive training unit; The operation deviation data is packaged and uploaded to the self-evolution knowledge base unit to enrich the case library;

[0048] Operation deviation degree formula: , in the formula: represents the operation deviation degree, with the unit of mm·s, which comprehensively reflects the space-time difference between the student and the expert operation, such as the cable joint crimping operation, is qualified;

[0049] represents the student operation trajectory vector, including three-dimensional position (unit: mm), force size (unit: N), and tool angle (unit: °), which is collected in real time by the 6DoF sensor of the AR glasses and the force sensor of the exoskeleton gloves, with a sampling rate of 100 Hz;

[0050] represents the expert holographic template trajectory vector, generated by recording the standard operation of experienced electricians, such as the force vector when crimping the cable is (0, 0, 35) N (axial force);

[0051] respectively represent the operation start time and the operation end time, in seconds, which are automatically identified by the system, such as the time interval from picking up the wire stripper to completing the insulation stripping;

[0052] represents the key node weight, dimensionless, which is set according to the operation risk level, such as cable terminal installation , control cabinet wiring , the higher the risk, the greater the weight;

[0053] Function: Accurate quantification of operation specification, when , the exoskeleton feedback device triggers a vibration reminder (the frequency increases with ), and generates a deviation analysis report, such as "crimping angle deviation 3°", which is uploaded to the self-evolution knowledge base unit for updating the fault tolerance threshold of the expert template.

[0054] Preferably, the self-evolution knowledge base unit is used for:

[0055] (1) Multi-source data integration and collaborative training: adopt distributed architecture to integrate multi-source information such as cross-domain fault chain cases, holographic operation deviation data and brain-machine cognitive map; through the federal learning framework, collaboratively train the data under the premise of protecting privacy, refine the operation specification update rules and fault mode recognition model;

[0056] Federal learning framework aggregation formula: , where: represents the global update model parameters, including fault recognition threshold, operation scoring standard, etc., such as the cable insulation resistance qualified threshold is updated from 800 ;

[0057] represents the local model parameters of the th factory area, which is generated by independent training of training data from each factory area, such as the insulation resistance threshold is lowered by 15% in A factory area due to humid environment;

[0058] represents the sample weight of the th factory area, which is positively related to the number of training personnel and the number of equipment types in the factory area, such as the factory area with 5 cable production lines , only 1 production line ;

[0059] This indicates the number of factories participating in federated learning. The initial pilot program will consist of 5 factories, which will be gradually expanded to the entire industry. Data privacy will be protected through encryption protocols.

[0060] Purpose: To aggregate industry experience without sharing the original data, and to update the data accordingly. Synchronize with the quantum-enhanced digital twin unit to correct the aging model of insulation materials, such as the humidity influence coefficient in coastal factories, and provide the latest fault association rules for the cross-domain fault chain generation unit, such as the protection logic of the new intelligent control cabinet.

[0061] (2) Knowledge synchronization and template iteration: The updated knowledge base is synchronized to the quantum-enhanced digital twin unit to optimize the physical model, providing the latest association rules for the cross-domain fault chain generation unit; based on the expert data update frequency of the holographic operation memory unit, the expert template iteration is automatically triggered to ensure that the training content always matches the industry best practices.

[0062] This invention also provides a digital training method based on safe production. Based on the above system, the specific steps of this method are as follows:

[0063] S1. Neural-device collaborative modeling: Microscopic modeling generates digital samples of cables and control cabinets, which are imported into the quantum-enhanced digital twin unit; neural baseline calibration collects expert data and establishes a cognitive model stored in the brain-computer collaborative cognitive training unit; virtual and real parameters are bound together with physical indicators (such as cable insulation resistance and control cabinet temperature) and cognitive indicators (such as attention duration and operation tremor frequency) to form adaptation rules that are synchronized to the cross-domain fault chain generation unit, providing a data and rule foundation for subsequent training scenarios;

[0064] S2, Fault Chain Immersion Training: The cross-domain fault chain generation unit calls the real-time parameters of the quantum-enhanced digital twin unit and the case generation scenario of the self-evolving knowledge base unit. The complexity depends on the initial assessment setting of the brain-computer collaborative cognitive training unit. During training, the brain-computer collaborative cognitive training unit monitors the cognitive state and regulates the holographic operation memory unit or the cross-domain fault chain generation unit to adjust the scenario.

[0065] S3. Memory Enhancement and Continuous Optimization: The holographic operation memory unit retrieves expert templates and combines them with fault chain scenario characteristics to enhance muscle memory through an exoskeleton; the self-evolving knowledge base unit summarizes data and updates the model through federated learning, synchronizing it to the quantum-enhanced digital twin unit and the cross-domain fault chain generation unit; the neural adaptation test results are fed back to step S1, forming a process from training to system iteration.

[0066] Preferably, the specific steps of neural-device collaborative modeling in step S1 are as follows:

[0067] S11, basic model construction and feature extraction: data fusion is completed in stages: in the micro modeling stage, digital sample sets of cable insulation materials and control cabinet components are generated through molecular dynamics simulation, and a physical model is constructed by importing a quantum enhanced digital twin unit; in the neural baseline calibration link, electroencephalogram and motion data of skilled technicians during operation are collected, neural activity features are extracted, and a cognitive benchmark model is established and stored in the brain-machine collaborative cognitive training unit;

[0068] S12, parameter association and scene constraint setting: the key parameters of the physical model (such as insulation strength threshold) are associated and mapped with the cognitive benchmark indicators (such as attention fluctuation range) to form dynamic adaptation rules; related data is synchronized to the cross-domain fault chain generation unit as a constraint condition for scene construction, realizing deep coupling between the physical model and the cognitive model;

[0069] Virtual-real parameter mapping coefficient formula: , wherein: represents the mapping coefficient, with a value of -1~1, the greater the absolute value, the stronger the correlation, for example, K=0.85 indicates that the cable temperature is strongly positively correlated with the student's attention;

[0070] represents the covariance of the physical parameter and the cognitive indicator , and represents the cable insulation resistance (unit: MΩ) or the control cabinet current (unit: A), represents the student's attention duration (unit: s) or the number of error operations;

[0071] represents the variance of the physical parameter , reflecting the fluctuation degree of the device state, for example, the insulation resistance variance of an old cable is 30% higher than that of a new cable;

[0072] represents the variance of the cognitive indicator , reflecting the difference in operation stability of different students, for example, the error number variance of a novice is 2 times higher than that of a skilled person;

[0073] Effect: establish a correlation model between device state and student cognition, for example, when =−0.7 (negative correlation), the system automatically extends the student's attention evaluation period when the insulation resistance decreases, and the correlation data is used for dynamic adjustment of the scene difficulty in the cross-domain fault chain generation unit, such as matching low cognitive load in high-risk scenarios.

[0074] Preferably, the specific steps of the fault chain immersion training in step S2 are as follows:

[0075] S21, scene generation and complexity adaptation: based on the basic model constructed in step S1, the cross-domain fault chain generation unit calls the real-time parameters of the quantum enhanced digital twin unit, combines the cases in the self-evolution knowledge base, and constructs a multi-link virtual scene; the scene complexity is set according to the initial cognitive evaluation result of the trainee fed back by the brain-machine collaborative cognitive training unit, to ensure that the training difficulty matches the ability of the trainee;

[0076] S22, real-time regulation and review analysis: the brain-machine collaborative cognitive training unit continuously monitors the state of the trainee, and triggers regulation when the cognition deviates from the benchmark: when the tension is too high, the holographic operation memory unit reduces the scene stimulation and adds prompts; when the attention is scattered, the cross-domain fault chain generation unit increases the fault branches; after training, the holographic review integrates the fault chain, operation trajectory and neural curve, to provide targeted direction for memory reinforcement. Preferably, the specific steps of memory reinforcement and continuous optimization in step S3 are as follows:

[0077] S31, operation reinforcement and data iteration: based on the training results in step S2, the holographic operation memory unit adjusts the force feedback parameters according to the weak points identified by the review, and retrieves the expert template to guide the trainee to repeat the training through the exoskeleton device. The self-evolution knowledge base unit summarizes the operation data and updates the fault rules and specifications through federated learning;

[0078] Feedback intensity adjustment formula:

[0079] ,

[0080] In the formula: represents the feedback intensity of the exoskeleton, which is applied by the air pressure module built in the glove (range 5~30N), to ensure that the trainee can clearly perceive the difference in operation intensity.

[0081] represents the reference feedback intensity, which is the average intensity of expert operation, such as the pressure of cable joint crimping , extracted from the expert template of the holographic operation memory unit;

[0082] represents the operation deviation (unit mm·s), which directly uses the calculation result of the holographic operation memory unit, and the higher the deviation, the higher the feedback intensity;

[0083] represents the scene complexity, which is the number of nodes contained in the fault chain, such as "broken cable → partial discharge → control cabinet tripping", which is output in real time by the cross-domain fault chain generation unit.

[0084] represents the scene sensitivity coefficient , through experimental calibration, ensure that the feedback intensity increase in complex scene is reasonable (avoid exceeding the limit of the student);

[0085] Effect: dynamically match the feedback intensity with the training needs, for example, in and high difficulty scenarios, will reach 1.5 times the baseline value, strengthen muscle memory training effect, feedback data for neural adaptation test link, optimize the stress adaptation threshold of the cognitive benchmark model;

[0086] S32, adaptive test and system iteration: introduce random interference factors, such as virtual equipment parameter fluctuation, carry out neural adaptation test, evaluate the operation stability of students in complex environment; the test results are fed back to step S1, the optimization of the cognitive benchmark model and the parameter binding rule, forming a closed-loop process from training implementation to system iteration, improving the training effect.

[0087] The beneficial effects of the present application are as follows:

[0088] 1. The micro-macro double-domain mapping is constructed by the quantum enhancement digital twin unit, so that the student can directly observe the microcosmic process such as the breaking of the molecular chain of the cable insulation layer and the evolution of the plasma of the control cabinet arc discharge, and is associated with the macroscopic fault performance; This multi-scale visual training solves the problem that the traditional system can only present the surface state of the equipment, helps the student to understand the formation mechanism of the potential risk, and improves the prediction ability of the hidden danger of high-voltage equipment from the root, avoiding the operation omission caused by insufficient risk awareness.

[0089] 2. The brain-computer collaborative cognitive training unit can monitor the neural activity and operation characteristics of the student in real time, and realize the precise regulation of cognitive load by adjusting the stress parameters (such as the intensity of the sound and light warning of the high-voltage equipment) of the virtual scene; Compared with the traditional standardized training, this mechanism can effectively alleviate the decision delay problem of the student in the real high-voltage operation, so that the training effect is closer to the actual working condition, and ensures that the student can maintain stable judgment and execution ability when facing high-risk operations such as 35kV cable pressure test.

[0090] 3. The cross-domain fault chain generation unit constructs a fault correlation graph based on a graph neural network, which can automatically generate chain reaction scenes such as "cable joint defect → partial discharge → insulation breakdown → control cabinet protection misoperation"; This dynamic evolution training content makes up for the defect of isolated fault scenes in traditional systems, so that students can master the conduction law of faults in actual production, improve the analysis and disposal ability of multi-linkage faults, and reduce the risk of expanded accidents caused by insufficient understanding of fault correlation. BRIEF DESCRIPTION OF DRAWINGS

[0091] Fig. 1 is a flowchart of the digital training system based on safety production of the present application;

[0092] Fig. 2 A flow chart of a digital training method based on safety production of the present application. DETAILED DESCRIPTION

[0093] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0094] As shown in Figs. 1-2 The embodiment of the present application provides a digital training system based on safety production, which comprises:

[0095] Quantum-enhanced digital twin unit: based on the physical properties of the equipment, a multi-scale digital twin body covering nanoscale (cable insulation layer molecular chain) to meter scale (overall structure of the control cabinet) is constructed, the microscopic changes of the cable insulation layer are tracked through quantum dot markers, and the macroscopic behavior of the control cabinet is associated with the microscopic mechanism; the high-speed interface synchronizes the real-time parameters to the cross-domain fault chain generation unit, supports dynamic scene construction, receives update data from the self-evolving knowledge base, continuously optimizes the model accuracy, and provides physical data support for the system;

[0096] Brain-machine collaborative cognitive training unit: integrate EEG and electromyography equipment, capture the neural activity and operation characteristics of the trainee, build a "cognition-action" model, when detecting cognitive abnormalities, send instructions to the holographic operation memory unit to trigger scene adaptation, encrypt cognitive data transmission to the self-evolving knowledge base unit, and also receive scene parameters from the cross-domain fault chain unit to dynamically adjust the monitoring threshold, realizing collaborative training of cognition and operation;

[0097] Cross-domain fault chain generation unit: based on real-time parameters and historical cases of the self-evolving knowledge base unit, a device correlation graph is constructed using a graph neural network, and a fault chain scene is generated based on trainee operation data and virtual device status, which is synchronized to the holographic operation memory unit for driving rendering, and the rules are output to the self-evolving knowledge base unit for iteration;

[0098] Holographic operation memory unit: record expert operations through light field imaging and generate holographic templates, after receiving scene parameters from the cross-domain fault chain generation unit, realize "same screen operation" of the trainee and the virtual expert; the AR device compares the operation deviation, and the exoskeleton gloves provide tactile guidance (vibration frequency 5~20Hz) at the part where the thumb and index finger contact the cable crimping tool, the strength and frequency are dynamically adjusted according to the trainee state of the brain-machine collaborative cognitive training unit, and the deviation data is uploaded to the self-evolving knowledge base unit;

[0099] Self-evolution knowledge base unit: integrate multi-source information including fault chain, holographic operation and brain-computer collaboration, refine rules and models through federated learning, update and synchronize to quantum-enhanced digital twin unit to optimize physical model, provide the latest rules for cross-domain fault chain generation unit, and trigger expert template iteration.

[0100] The quantum-enhanced digital twin unit is used for:

[0101] (1) Multi-scale modeling and microscopic mechanism restoration: Based on the physical properties of intelligent power equipment, a multi-scale digital twin body is constructed. Quantum dot marking algorithm is used for cable insulation layer modeling to track virtual quantum dot trajectories and dynamically present microscopic changes such as molecular chain rupture and crystal zone ablation under the action of temperature and stress. Intelligent control cabinet is associated with macro behaviors such as circuit breaker action and microscopic mechanisms such as electron transition through electromagnetic simulation module to realize multi-dimensional restoration of equipment state and provide physical basis for subsequent scene construction;

[0102] Quantum dot marking algorithm steps: 1-5 nm virtual quantum dots are implanted in the digital twin model based on the molecular structure of cable insulation materials; the binding energy of quantum dots and molecular chains is calculated through molecular dynamics simulation (threshold value ≥ 0.2 eV); 3D coordinates of quantum dots are sampled at a frequency of 10 kHz, and Kalman filter is used to smooth the trajectory and identify molecular chain rupture (trajectory mutation ≥ 5 μm / s);

[0103] In the quantum dot marking algorithm, the Kalman filter parameters are set as follows: filter gain K = 0.3, iteration number N = 50, and molecular chain rupture is determined when the trajectory mutation threshold is ≥ 5 μm / s;

[0104] Quantum dot diffusion coefficient formula:

[0105] ,

[0106] In the formula: represents the quantum dot diffusion coefficient, with a unit of is a core index describing the activity of virtual molecular motion in cable insulation layer; in the digital twin scene, its value directly corresponds to the looseness of the microstructure of the insulation material, and the larger the value, the higher the risk of molecular chain rupture;

[0107] represents the initial diffusion coefficient, whose reference value is determined by the type of cable insulation material, such as for cross-linked polyethylene cable, for rubber insulated cable, which is obtained through material laboratory measurement and pre-recorded into the system;

[0108] represents the diffusion activation energy, unit: eV, characterizes the energy threshold required for the molecular chain of insulating materials to break; for 10 kV cables, its value is usually between 0.3 and 0.5 eV, and is dynamically corrected according to the degree of material aging (decreased by 15% to 20% for aged cables compared to new cables);

[0109] represents the Boltzmann constant , used to convert temperature parameters into energy units, ensuring dimensional consistency of the formula;

[0110] represents the virtual environment temperature, unit: K, synchronized with the real-time temperature of the cable production line (range 293-353 K), collected from the temperature control module of the extruder through the device data interaction unit;

[0111] represents the mechanical stress, unit: MPa, corresponding to the tension during cable laying, such as 8-12 MPa for overhead cables and 5-8 MPa for underground direct-buried cables, generated by converting the laying angle and traction force parameters in the virtual scene;

[0112] represents the stress sensitivity coefficient, dimensionless, calibrated through 1000 cable stretching experiments (polyethylene material takes , reflecting the acceleration effect of mechanical force on molecular diffusion;

[0113] Effect: Quantify the coupling effect of temperature and mechanical stress on microscopic motion, and the calculation results are used to drive the insulating layer tree branch growth animation in the digital twin unit trigger partial discharge simulation (visual: AR device renders a blue arc dynamic; auditory: earphones play 1000-2000 Hz buzzing sound; tactile: weak electric stimulation feedback at the wrist of the exoskeleton), and push real-time data to the cross-domain fault chain generation unit as the starting condition for the "insulation failure → control cabinet tripping" fault chain;

[0114] (2) Data linkage and model iterative optimization: The unit is equipped with a high-speed interface that synchronizes real-time parameters such as cable insulation resistance and control cabinet temperature field to the cross-domain fault chain generation unit, supporting dynamic fault scenario construction; at the same time, it receives device characteristic update data from the self-evolution knowledge base unit, continuously optimizes the physical precision of the twin model, and ensures that the virtual scene is highly consistent with the actual device characteristics.

[0115] Among them, the brain-computer collaborative cognitive training unit is used to:

[0116] (1) Multi-dimensional state monitoring and model construction: integrate EEG acquisition and electromyographic sensing equipment to capture the neural activity and operation characteristics of students in real time; the EEG module identifies cognitive states such as attention and emotion through frontal lobe electrodes; the electromyographic equipment judges the stability of hand operation, and the core algorithm fuses data to construct a "cognition-action" correlation model to accurately identify cognitive load abnormalities in high-voltage operation and provide a basis for scene regulation;

[0117] The EEG device uses an 8-channel frontal lobe electrode array with a sampling frequency of 250Hz (compliant with IEC60601-2-26 standard), and the electromyographic sensor has a sampling rate of 1kHz;

[0118] Cognitive load index formula:

[0119] ,

[0120] In the formula: Cognitive load index, value range 0~1, 0.7 is the critical value, in high-voltage cable wiring training, Indicates that the student has the risk of operation failure, such as accidentally touching the live terminal;

[0121] EEG represents the amplitude ratio of theta wave (4~8Hz) and beta wave (13~30Hz), and the increase in theta wave reflects the anxiety state, such as when facing a 35kV device, the increase in beta wave reflects the concentration of attention, and the frontal lobe electrode array of the EEG earphone is collected (sampling rate ;

[0122] EEG represents the real-time electromyographic signal amplitude, with units of μV, collected by the wrist electromyographic sensor, reflecting the tremor degree of hand operation (normal operation , nervous ;

[0123] EEG represents the maximum electromyographic signal amplitude, with units of μV, obtained by initial testing of students, such as the electromyographic value when holding the cable stripping pliers with full force, as a benchmark for individual operation limits;

[0124] EEG represents the weight coefficient , calibrated based on the data of 100 electrician training, and the weight is higher because the neural signal has a more significant impact on operation safety;

[0125] Effect: Real-time assessment of students' physical and mental state, when >0.7, the system sends instructions to the holographic operation memory unit, reduces the sound and light intensity of the high-voltage arc in the virtual scene (by 30%), and triggers expert operation holographic prompts, while​​ The time-series data is encrypted and uploaded to a self-evolving knowledge base to optimize the difficulty curve of personalized training programs;

[0126] (2) Data interaction and dynamic adjustment adaptation: When a cognitive abnormality is detected, an adjustment instruction (format: [Scene ID, adjustment type, intensity value], such as [FD-001, reduced sound and light, 30%]) is generated and sent to the holographic operation memory unit to trigger scene adaptation. At the same time, the encrypted student cognitive data is transmitted to the self-evolving knowledge base unit to support personalized solution optimization; the scene complexity parameters are received from the cross-domain fault chain generation unit to dynamically adjust the cognitive monitoring threshold and achieve accurate matching between the training scene and the student's state.

[0127] The cross-domain fault chain generation unit is used for:

[0128] (1) Construction of the association graph and scenario triggering: Based on the physical parameters of the quantum-enhanced digital twin unit and the historical cases of the self-evolving knowledge base, the device association graph is constructed through the graph neural network; the graph includes cable production parameters and control cabinet components, and the node weights are dynamically adjusted according to the linkage relationship. When training is started, the system triggers a chain failure scenario such as cable laying deviation based on the trainee's operation data and the virtual equipment status.

[0129] The graph neural network node features include insulation resistance and temperature for cable nodes, and current and circuit breaker status for control cabinet nodes. The network structure is defined as follows: input layer (128-dimensional features) → 2-layer GCN (64 / 32-dimensional hidden layers) → output layer (fault association weights). The training data consists of 10,000 historical fault chains, and the loss function is cross-entropy (convergence threshold ≤ 0.01).

[0130] The hidden layer of the graph neural network uses the ReLU activation function, the cross-entropy loss function convergence threshold is ≤0.01, and the training samples are 10,000 historical fault chain data.

[0131] Fault association weight formula:

[0132] ,

[0133] In the formula: Represents device node To the node The fault association weight, with a value of 0 to 1, is used to quantify the fault propagation probability (e.g., show After the malfunction There is an 80% chance of triggering this.

[0134] Indicating historical data trigger The number of failures, derived from the failure case library of the self-evolution knowledge base, such as the number of records of "cable extruder pressure anomaly (i) -> insulation layer off-center (j)";

[0135] The physical parameters of the node , The cable parameters, such as extruder pressure MPa, The control cabinet parameters, such as PLC temperature alarm threshold ;

[0136] The parameter reference value, the rated parameter of the equipment under normal operation, such as the rated pressure of the extruder MPa, the rated temperature of the control cabinet , determined by the equipment manual;

[0137] Function: Calculate the device correlation strength through graph neural network, for example, when = 0.75, the system preferentially generates a fault chain of "extruder pressure exceeds standard -> cable resistance anomaly -> control cabinet over-current protection action", the weight matrix is synchronized to the holographic operation memory unit, driving the linkage change of the device state in the virtual scene (such as the flashing of the control cabinet indicator light with the fluctuation of the cable parameter);

[0138] (2) Data synchronization and rule iteration: the generated dynamic fault chain data is synchronized to the holographic operation memory unit to drive scene rendering, and the reverse output fault evolution rules are output to the self-evolution knowledge base unit, when the cumulative verification number of a single rule reaches 50 times and the accuracy is ≥ 90%, the rule iteration update is triggered; through continuous reception of multi-source data, the correlation graph is optimized to ensure the coherence and timeliness of the fault scene, and to provide complex scenes close to reality for training.

[0139] Among them, the holographic operation memory unit is used for:

[0140] (1) Expert template construction and on-screen operation: record expert operation data through light field imaging, form interactive holographic templates through three-dimensional reconstruction, receive dynamic fault chain data of the cross-domain fault chain generation unit during training, call matching templates, and realize "on-screen operation" of students and virtual experts through holographic projection; AR devices capture student actions in real time and compare them with templates to generate deviation reports;

[0141] (2) Tactile guidance and data feedback: the exoskeleton feedback device provides tactile guidance at key nodes such as cable crimping according to the comparison results, and the strength and frequency are dynamically adjusted according to the student state of the brain-computer collaborative cognitive training unit; operation deviation data is packaged and uploaded to the self-evolution knowledge base unit to enrich the case library;

[0142] The haptic guidance intensity is adjusted based on a feedback intensity adjustment formula (F) through a PWM signal to control the air pressure pump (range 0-100 kPa); when the deviation δ is greater than 100 mm·s, the vibration frequency is linearly increased from 5 Hz to 20 Hz (step size 1 Hz / 10 mm·s);

[0143] The exoskeleton haptic guidance parameters: the initial frequency of the cable crimping scene is 10 Hz, and the frequency is increased by 2 Hz (maximum 20 Hz) for every 20 mm·s when δ is greater than 100 mm·s; the initial frequency of the control cabinet wiring scene is 8 Hz (maximum 15 Hz);

[0144] The operation deviation degree formula is: , wherein: represents the operation deviation degree, with a unit of mm·s, which comprehensively reflects the spatial and temporal differences between the student and the expert operations, such as the cable joint crimping operation, is qualified;

[0145] represents the student operation trajectory vector, which includes three-dimensional position (unit: mm), force (unit: N), and tool angle (unit: °), and is collected in real time by the 6DoF sensor of the AR glasses and the force sensor of the exoskeleton gloves, with a sampling rate of 100 Hz;

[0146] represents the expert holographic template trajectory vector, which is generated by recording the standard operation of a senior electrician, such as the force vector (0, 0, 35) N (axial force) during cable crimping;

[0147] The trajectory vector and have three-dimensional position coordinates with the center point of the cable joint in the virtual scene as the origin, and the Z-axis as the cable axial direction, with a unit conversion accuracy of ±0.1 mm;

[0148] represents the operation start time and operation end time, respectively, with a unit of s, which is automatically identified by the system, such as the time interval from picking up the wire stripper to completing the insulation layer stripping;

[0149] represents the key node weight, which is dimensionless, and is set according to the operation risk level, with the cable terminal installation , control cabinet wiring , and the higher the risk, the greater the weight;

[0150] Effect: accurately quantifying the operation specification, when , the exoskeleton feedback device triggers a vibration reminder (the frequency increases with ), and generates a deviation analysis report, such as "crimping angle deviation 3°", which is uploaded to the self-evolution knowledge base unit for updating the fault tolerance threshold of the expert template.

[0151] Wherein, the self-evolution knowledge base unit is used for:

[0152] (1) Multi-source data integration and collaborative training: Adopting a distributed architecture to integrate multi-source information such as cross-domain fault chain cases, holographic operation deviation data, and brain-machine cognitive maps; Through a federated learning framework, collaboratively train the data under the premise of protecting privacy, refine the operation specification update rules and fault mode recognition models;

[0153] The federated learning framework trains the local model with 5000 data (optimizer Adam, learning rate 0.001); The FedAvg algorithm is used, and the model parameters are transmitted through homomorphic encryption; Every 100 new data triggers aggregation, and the convergence condition is that the difference between the global model parameters of the adjacent two times is less than or equal to 1e-5;

[0154] The federated learning adopts the Paillier homomorphic encryption protocol, and the public key length is 2048 bits. Every 100 new operation data triggers global model aggregation;

[0155] The aggregation formula of the federated learning framework is: , wherein: represents the global updated model parameters, including fault recognition thresholds, operation scoring standards, etc., such as the qualified threshold of cable insulation resistance from to 800 ;

[0156] represents the local model parameters of the th factory area, which is generated by independent training of the training data of each factory area, such as the A factory area reducing the insulation resistance threshold by 15% due to the humid environment;

[0157] represents the sample weight of the th factory area, which is positively related to the number of training personnel and the number of equipment types in the factory area, such as the factory area with 5 cable production lines , and the factory area with only 1 production line ;

[0158] represents the number of factory areas participating in federated learning, initially 5, and gradually expanded to the entire industry, realizing data privacy protection through encryption protocols;

[0159] Function: Aggregate industry experience without sharing raw data, and synchronize the updated to the quantum-enhanced digital twin unit to correct the insulation material aging model, such as the humidity influence coefficient of coastal factory areas, and provide the latest fault correlation rules for the cross-domain fault chain generation unit, such as the protection logic of new intelligent control cabinets;

[0160] (2) Knowledge synchronization and template iteration: The updated knowledge base is synchronized to the quantum-enhanced digital twin unit to optimize the physical model, and the latest association rules are provided to the cross-domain fault chain generation unit; according to the expert data update frequency of the holographic operation memory unit, the expert template iteration is automatically triggered to ensure that the training content always matches the industry best practices;

[0161] Expert template iteration trigger condition: automatically start iteration when 50 new expert operation data are accumulated and δ < 30 mm・s (10% better than historical template).

[0162] The embodiment of the application also provides a digital training method based on safety production, based on the above-mentioned system, the specific steps of the method are as follows:

[0163] S1, neural-device collaborative modeling: micro-modeling generates digital samples of cables and control cabinets, and imports the quantum-enhanced digital twin unit; neural baseline calibration collects expert data, establishes a cognitive model, and stores it in the brain-machine collaborative cognitive training unit; virtual-real parameter binding associates physical and cognitive indicators to form adaptive rules and synchronize to the cross-domain fault chain generation unit to provide data and rule basis for subsequent training scenarios;

[0164] S2, fault chain immersion training: the cross-domain fault chain generation unit calls the real-time parameters of the quantum-enhanced digital twin unit and the cases generated by the self-evolving knowledge base unit to generate scenarios, and the complexity is set according to the initial evaluation of the brain-machine collaborative cognitive training unit; during training, the brain-machine collaborative cognitive training unit monitors the cognitive state and adjusts the scenario by regulating the holographic operation memory unit or the cross-domain fault chain generation unit;

[0165] S3, memory reinforcement and continuous optimization: the holographic operation memory unit retrieves the expert template, combines the fault chain scenario characteristics, and strengthens muscle memory through the exoskeleton; the self-evolving knowledge base unit aggregates data and updates the model through federated learning, synchronizes to the quantum-enhanced digital twin unit and the cross-domain fault chain generation unit, and the neural adaptation test results are fed back to step S1 to form a process from training to system iteration.

[0166] Example one: a digital training scene case of improper 35kV cable crimping operation causing failure

[0167] S1: neural-device collaborative modeling

[0168] The quantum-enhanced digital twin unit generates a digital sample of the 35kV cable insulation layer (cross-linked polyethylene material) and the control cabinet, and tracks the molecular chain state of the insulation layer (initial diffusion coefficient D0=1.2×10−3μm2 / s, stable molecular chain) through quantum dot markers; the brain-computer collaborative cognitive training unit collects expert crimping operation electroencephalogram (theta / beta=0.3 when concentrating) and electromyogram (stable operation EMG=30μV) data, and establishes a cognitive benchmark model (CLI=0.32); after virtual-real parameter binding, the correlation rule between "cable crimping force and student attention" (K=-0.6, negative correlation) is synchronized to the cross-domain fault chain generation unit.

[0169] S2: Fault chain immersion training

[0170] The cross-domain fault chain generation unit calls the twin data and knowledge base cases to generate a fault chain of "insufficient crimping force → insulation layer damage → partial discharge → control cabinet alarm" (correlation weight of improper crimping partial discharge), and the scene complexity is set to S=3 according to the initial cognitive assessment of the student (CLI=0.5).

[0171]

[0172] During training, the student's crimping force is too small, and the quantum-enhanced digital twin unit monitors that the molecular chain of the insulation layer is loose (D=5.2×10−3μm2 / s, triggering partial discharge simulation (vision: AR device renders a dynamic blue arc; hearing: earphones play a 1000~2000Hz buzzing sound; touch: weak electric stimulation feedback at the wrist of the exoskeleton)); the brain-computer collaborative cognitive training unit finds that the student is nervous due to the sound and light stimulation of the discharge (CLI=0.75, exceeding the critical value), immediately instructs the holographic unit to reduce the sound and light by 30%, and projects the expert crimping holographic prompt; after the student's attention is dispersed, the cross-domain fault chain generation unit adds a "control cabinet misoperation" branch (S=4) to strengthen training.

[0173] S3: Memory reinforcement and optimization

[0174] The holographic operation memory unit compares the student's actions and calculates the operation deviation δ=110mm·s (exceeding the qualified threshold), and the exoskeleton device provides a 30N tactile guide according to the deviation and scene complexity (20N stronger than the benchmark force); the self-evolution knowledge base unit summarizes the operation data, updates the fault rules through federated learning (the crimping force threshold is lowered by 5%), and synchronizes them to the quantum-enhanced digital twin unit to optimize the insulation layer model; after the neural adaptation test, the results are fed back to step S1 to fine-tune the cognitive benchmark model.

[0175] The specific steps of the neural-device collaborative modeling in step S1 are as follows:

[0176] ​S11. Basic Model Construction and Feature Extraction: Data fusion is completed in stages: In the microscopic modeling stage, digital sample sets of cable insulation materials and control cabinet components are generated through molecular dynamics simulation and imported into the quantum-enhanced digital twin unit to construct a physical model; In the neural baseline calibration stage, EEG and motion data of senior technicians during operation are collected, neural activity features are extracted, and a cognitive benchmark model is established and stored in the brain-computer collaborative cognitive training unit.

[0177] S12. Parameter Association and Scene Constraint Setting: Key parameters of the physical model (such as insulation strength threshold) are associated and mapped with cognitive benchmark indicators (such as attention fluctuation range) to form dynamic adaptation rules; relevant data are synchronized to the cross-domain fault chain generation unit as constraints for scene construction to achieve deep coupling between the physical model and the cognitive model.

[0178] Formula for mapping coefficients between real and virtual parameters: In the formula: This represents the mapping coefficient, ranging from -1 to 1. The larger the absolute value, the stronger the correlation. For example, K=0.85 indicates a strong positive correlation between cable temperature and student attention.

[0179] Representing physical parameters With cognitive indicators covariance, The current is either the cable insulation resistance (in MΩ) or the control cabinet current (in A). This refers to the duration of a student's attention (in seconds) or the number of incorrect actions.

[0180] Representing physical parameters The variance reflects the degree of fluctuation in equipment condition; for example, the variance of insulation resistance of old cables is 30% higher than that of new cables.

[0181] Indicators of cognitive ability The variance reflects the difference in operational stability among different trainees; for example, the variance of the number of errors made by a novice is twice as high as that of an experienced trainee.

[0182] Purpose: To establish a correlation model between equipment status and learner cognition, for example, when... When the correlation is -0.7 (negative correlation), the system automatically extends the student's attention assessment cycle when the insulation resistance decreases. The correlated data is used to dynamically adjust the scenario difficulty of the cross-domain fault chain generation unit, such as matching high-risk scenarios with low cognitive load.

[0183] The specific steps of fault chain immersion training in step S2 are as follows:

[0184] S21, scene generation and complexity adaptation: based on the basic model constructed in step S1, the cross-domain fault chain generation unit calls the real-time parameters of the quantum enhanced digital twin unit, combines the cases in the self-evolution knowledge base, and constructs a multi-link virtual scene; the scene complexity is set according to the initial cognitive evaluation result of the trainee fed back by the brain-machine collaborative cognitive training unit, to ensure that the training difficulty matches the ability of the trainee;

[0185] S22, real-time regulation and review analysis: the brain-machine collaborative cognitive training unit continuously monitors the state of the trainee, and triggers regulation when the cognition deviates from the benchmark: if the trainee is too nervous, the holographic operation memory unit reduces the scene stimulation and adds prompts; if the trainee is distracted, the cross-domain fault chain generation unit increases the fault branches; after training, the holographic review integrates the fault chain, operation trajectory and neural curve to provide targeted direction for memory reinforcement.

[0186] Among them, the specific steps of memory reinforcement and continuous optimization in step S3 are as follows:

[0187] S31, operation reinforcement and data iteration: based on the training results in step S2, the holographic operation memory unit adjusts the force feedback parameters according to the weak points identified by the review, and retrieves the expert template to guide the trainee to repeat the training through the exoskeleton device. The self-evolution knowledge base unit summarizes the operation data and updates the fault rules and specifications through federated learning;

[0188] Feedback intensity adjustment formula:

[0189] ,

[0190] In the formula: represents the feedback intensity of the exoskeleton, which is applied by the air pressure module built-in the glove (range 5~30N), to ensure that the trainee can clearly perceive the difference in operation intensity.

[0191] represents the benchmark feedback intensity, which is the average intensity of expert operation, such as the pressure of cable joint crimping , extracted from the expert template of the holographic operation memory unit;

[0192] represents the operation deviation (unit mm·s), which directly uses the calculation result of the holographic operation memory unit, and the higher the deviation, the higher the feedback intensity;

[0193] represents the scene complexity, which is the number of nodes contained in the fault chain, such as “cable damage → partial discharge → control cabinet tripping”, which is output in real time by the cross-domain fault chain generation unit.

[0194] represents the scene sensitivity coefficient , through experimental calibration, ensure that the feedback intensity increase in complex scenarios is reasonable (avoid exceeding the limit of the trainee);

[0195] Effect: dynamically match feedback intensity with training needs, for example, in and high difficulty scenarios, will reach 1.5 times the baseline value, strengthen muscle memory training effect, feedback data for neural adaptation test link, optimize the stress adaptation threshold of cognitive benchmark model;

[0196] S32, adaptive test and system iteration: introduce random interference factors, such as virtual equipment parameter fluctuation, carry out neural adaptation test, evaluate the operation stability of trainees in complex environment; the test results are fed back to step S1, the cognitive benchmark model and the parameter binding rules are optimized, forming a closed-loop process from training implementation to system iteration, and improving the training effect.

[0197] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0198] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A safety production-based digital training system, characterized in that: The system comprises: Quantum-enhanced digital twin unit: based on the physical properties of intelligent power equipment, a multi-scale digital twin body is constructed, quantum dot marking algorithm is used for cable insulation layer modeling, virtual quantum dot trajectory is tracked, and micro changes including temperature, stress, molecular chain rupture under stress, and crystal region ablation are dynamically presented; the intelligent control cabinet is associated with the macro behavior including the action of the circuit breaker and the micro mechanism of the electron transition through the electromagnetic simulation module; the unit is built-in high-speed interface, which synchronizes real-time parameters including cable insulation resistance and control cabinet temperature field to the cross-domain fault chain generation unit, and receives equipment characteristic update data from the self-evolution knowledge base unit; Brain-computer collaborative cognitive training unit: integrate EEG and electromyographic equipment, capture the neural activity and operation characteristics of the student, build a cognitive-action model, when cognitive abnormalities are detected, send instructions to the holographic operation memory unit, encrypt the cognitive data and transmit it to the self-evolution knowledge base unit, and dynamically adjust the monitoring threshold; Cross-domain fault chain generation unit: based on real-time parameters, use graph neural network to construct a device association graph, generate a fault chain scene according to the student operation data and the virtual device state, synchronize the scene parameters to the holographic operation memory unit, and output the rules to the self-evolution knowledge base unit; Holographic operation memory unit: record expert operations and generate holographic templates through light field imaging, after receiving the scene parameters from the cross-domain fault chain generation unit, realize the on-screen operation of the student and the virtual expert, the AR device compares the operation deviation, the exoskeleton device provides tactile guidance, and the strength and frequency are adjusted according to the student state of the brain-computer collaborative cognitive training unit; Self-evolution knowledge base unit: integrate multi-source information including fault chain, holographic operation and brain-computer collaboration, refine rules and models through federated learning, and update them to the quantum-enhanced digital twin unit to optimize the physical model and trigger expert template iteration.

2. The digital training system based on safety production according to claim 1, characterized in that: The brain-computer collaborative cognitive training unit is used for: (1) Multi-dimensional state monitoring and model construction: integrate EEG acquisition and electromyographic sensing equipment, real-time capture student neural activity and operation characteristics, EEG module identifies cognitive states including attention and emotion through frontal lobe electrodes; electromyographic equipment judges hand operation stability, and fuses data to build a cognitive-action correlation model; (2) Data interaction and dynamic regulation adaptation: when cognitive abnormalities are detected, generate regulation instructions and send them to the holographic operation memory unit to trigger scene adaptation, and at the same time, encrypted student cognitive data is transmitted to the self-evolution knowledge base unit, scene complexity parameters are received from the cross-domain fault chain generation unit, and cognitive monitoring threshold is dynamically adjusted.

3. The digital training system based on safety production according to claim 2, characterized in that: The cross-domain fault chain generation unit is used for: (1) Association graph construction and scene triggering: construct a device association graph through graph neural network, the graph includes cable production parameters and control cabinet components, node weights are dynamically adjusted according to the linkage relationship, and when training is started, according to the student operation data and the virtual device state, trigger a chain fault scene, including a chain fault caused by cable laying deviation; (2) Data synchronization and rule iteration: the generated dynamic fault chain data is synchronized to the holographic operation memory unit to drive scene rendering, and at the same time, the fault evolution rules are output to the self-evolution knowledge base unit in reverse.

4. The digital training system based on safety production according to claim 3, characterized in that: The holographic operation memory unit is used for: (1) Expert template construction and on-screen operation: Record expert operation data through light field imaging, form interactive holographic templates through three-dimensional reconstruction, receive dynamic fault chain data from the cross-domain fault chain generation unit during training, call matching templates, and realize student and virtual expert on-screen operation through holographic projection. The AR device captures student actions in real time and compares them with the template to generate a deviation report; (2) Haptic guidance and data feedback: The exoskeleton feedback device provides haptic guidance at key nodes including cable crimping based on the comparison results, and the intensity and frequency are dynamically adjusted according to the student state of the brain-computer collaborative cognitive training unit.

5. The digital training system based on safety production according to claim 4, characterized in that: The self-evolution knowledge base unit is used for: (1) Multi-source data integration and collaborative training: A distributed architecture is used to integrate multi-source information including cross-domain fault chain cases, holographic operation deviation data, and brain-computer cognitive maps. Through a federated learning framework, the data is collaboratively trained to refine the operation specification update rules and fault mode recognition models; (2) Knowledge synchronization and template iteration: The updated knowledge base is synchronized to the quantum-enhanced digital twin unit to optimize the physical model, and the expert template iteration is automatically triggered according to the expert data update frequency of the holographic operation memory unit.

6. A method of digital training based on safety production, based on the system of claim 5, characterized in that: The specific steps of the method are as follows: S1, Neuro-device collaborative modeling: Micro-modeling generates digital samples of cables and control cabinets, which are imported into the quantum-enhanced digital twin unit; Neuro-baseline calibration collects expert data to establish a cognitive model in the brain-computer collaborative cognitive training unit; Virtual-real parameter binding associates physical and cognitive indicators to form adaptive rules synchronized to the cross-domain fault chain generation unit; S2, Fault chain immersive training: The cross-domain fault chain generation unit calls the real-time parameters of the quantum-enhanced digital twin unit and the cases of the self-evolution knowledge base unit to generate scenarios, and the complexity is set according to the initial assessment of the brain-computer collaborative cognitive training unit. The brain-computer collaborative cognitive training unit monitors the cognitive state during training and triggers regulation when the cognitive state deviates from the baseline; S3, Memory reinforcement and continuous optimization: The holographic operation memory unit retrieves expert templates and combines fault chain scenario characteristics to reinforce muscle memory through exoskeletons; The self-evolution knowledge base unit aggregates data and updates models through federated learning, which are synchronized to the quantum-enhanced digital twin unit and the cross-domain fault chain generation unit. The results of the neuro-adaptation test are fed back to step S1.

7. The digital training method based on safety production according to claim 6, characterized in that: The specific steps of the neuro-device collaborative modeling in step S1 are as follows: S11, Basic model construction and feature extraction: Data fusion is completed in stages: In the micro-modeling stage, molecular dynamics simulation is used to generate digital sample sets of cable insulation materials and control cabinet components, which are imported into the quantum-enhanced digital twin unit to construct a physical model; In the neuro-baseline calibration stage, EEG and action data during expert operation are collected to establish a cognitive baseline model stored in the brain-computer collaborative cognitive training unit; S12, Parameter association and scenario constraint setting: Map key parameters of the physical model to cognitive baseline indicators to form dynamic adaptive rules; Related data is synchronized to the cross-domain fault chain generation unit as constraint conditions for scenario construction.

8. The digital training method based on safety production according to claim 7, characterized in that: The specific steps of the fault chain immersive training in step S2 are as follows: S21, scene generation and complexity adaptation: based on the basic model constructed in step S1, the cross-domain fault chain generation unit calls the real-time parameters of the quantum enhanced digital twin unit, combines the cases in the self-evolution knowledge base, constructs a virtual scene, and sets the scene complexity according to the initial cognitive evaluation results of the students fed back by the brain-computer collaborative cognitive training unit; S22, real-time regulation and review analysis: the brain-computer collaborative cognitive training unit continuously monitors the state of the students, and triggers regulation when the cognition deviates from the benchmark: when the tension is too high, the holographic operation memory unit reduces the scene stimulation and adds prompts; when the attention is scattered, the cross-domain fault chain generation unit increases the fault branches.

9. The digital training method based on safety production according to claim 8, characterized in that: The specific steps of memory reinforcement and continuous optimization in step S3 are as follows: S31, operation reinforcement and data iteration: based on the training results of step S2, the holographic operation memory unit calls the expert template to guide the students to repeat the training through the exoskeleton device, and the force feedback parameters are adjusted according to the characteristics of the cross-domain fault chain scene; The self-evolution knowledge base unit summarizes the operation data, and updates the fault rules and specifications through federated learning; S32, adaptive test and system iteration: random interference factors are introduced, including virtual device parameter fluctuation, neural adaptation test is carried out, the operation stability of students in complex environment is evaluated, and the test results are fed back to step S1.

Images