A digital cloud platform-based security practical training method and system

Abstract

The application relates to the technical field of safety practical training base management, and discloses a safety practical training method and system based on a digital cloud platform, which comprises the following steps: constructing a cognitive digital twin body which is synchronous with offline equipment; a cognitive confrontation engine generates virtual disturbance according to a cognitive path atlas, injects the twin body to deduce state changes and maps back to offline equipment, and records the operation sequence of students at the same time; the sequence is converted into individual cognitive paths and aggregated to the cognitive path atlas; the atlas is analyzed to identify group behavior patterns, and the results are used to drive the engine to generate new disturbance, forming a self-adaptive training closed loop; the system comprises offline practical training equipment, a central processing device and a terminal management device. Through self-adaptive training and deep quantitative analysis, the application overcomes the defects of traditional training staticity and shallow evaluation, can significantly improve the cognitive and decision-making ability of students under complex abnormal conditions, and enhances the pertinence and depth of training.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safety training base management, and particularly relates to a safety training method and system based on a digital cloud platform. BACKGROUND

[0002] In high-risk industries such as electric power and chemical industry, the emergency disposal ability of operators is the key to safety guarantee. Traditional offline training is costly and has limited opportunities, and more importantly, it cannot safely reproduce various high-risk and unexpected fault scenarios, resulting in a serious lack of training coverage and depth.

[0003] To address the above challenges, digital technologies such as virtual reality (VR) and digital twin are being applied in existing technologies to build safety training systems. These systems establish virtual models corresponding to physical devices in computers, allowing trainees to practice operations and simulate fault disposal in a virtual environment. This approach has made significant progress in reducing physical risks and equipment wear and tear in training.

[0004] However, existing digital training systems still have technical bottlenecks. Their core driving force is static training scripts, and they lack the ability to dynamically generate adversarial scenarios based on the actual level of trainees. Secondly, their evaluation system usually stops at simple judgments of right or wrong operation results, and cannot go deep into quantitative analysis of cognitive factors such as decision-making logic and reaction time in the disposal process, so it is difficult to reveal and correct the trainees' fixed thinking.

[0005] Therefore, the present application proposes a safety training method and system based on a digital cloud platform to solve the deficiencies of existing technologies. SUMMARY

[0006] The purpose of the present application is to provide a safety training method and system based on a digital cloud platform, which solves the problem of lack of real-time interaction and individualization in existing safety training, subjective evaluation, and inability to conduct in-depth analysis of group behavior.

[0007] To solve the above technical problems, the present application provides a safety training method and system based on a digital cloud platform.

[0008] The first aspect of the present application provides a safety training method based on a digital cloud platform, comprising the following steps:

[0009] A cognitive digital twin corresponding to the offline training device is constructed by a central processing device, and the operating parameters of the offline training device are synchronized in real time to the cognitive digital twin based on a sensor array on the offline training device;

[0010] Based on the cognitive confrontation engine in the central processing device, the cognitive path atlas is queried, virtual disturbance is dynamically generated and injected into the cognitive digital twin, the cognitive digital twin performs state deduction according to preset physical rules, and the deduced state changes are mapped to the offline practical training equipment through a content synchronization interface, and the cognitive confrontation engine records the operation sequence of the student for the offline practical training equipment;

[0011] Based on the cognitive confrontation engine, the operation sequence is converted into an individual cognitive path representing the decision-making process of the student.

[0012] The individual cognitive path is aggregated into the cognitive path atlas, and the cognitive path atlas is analyzed, and the analysis result is output, which is used to drive the cognitive confrontation engine to generate the virtual disturbance in the new training.

[0013] In one specific embodiment, the step of converting the operation sequence into an individual cognitive path is specifically converting the operation sequence into a weighted directed graph by the cognitive confrontation engine.

[0014] Wherein, the operation sequence is a time-ordered event tuple sequence In the formula, a i is the ith action performed by the student; t i is the timestamp of the action; o i is the object of the action; and k is the total number of operations performed by the student in one training.

[0015] The form of the weighted directed graph as an individual cognitive path is G ind =(V,E,W). In the formula, V is a node set; E is an edge set; and W is a weight set. Each node v i in the node set V corresponds to an event (a i ,t i ,o i ) in the operation sequence, each edge e i =(v i-1 ,v i ) in the edge set E represents the order flow between events, and is associated with a weight vector W i in the weight set W.

[0016] In another specific embodiment, the calculation formula of the weight vector w i is:

[0017] w i =[Δt i ,C(a i ),F(o i-1 )];

[0018] where Δt i = t i -t i-1 , denotes the decision reaction time; C(a i ) is the quantized value of the complexity of action a i ; F(o i-1 ) is the quantized value of the state feature feedback of the object o i-1 before the action is performed. The weight vector quantitatively represents the cognitive factors of the learner, such as the time consumption of the action, the complexity of the operation, and the observation of the state of the device.

[0019] In an optional embodiment, after the step of converting the operation sequence into individual cognitive paths, further comprising: the cognitive confrontation engine quantitatively evaluates the decision-making process of the learner by graph matching calculation between the individual cognitive paths and the expert paths extracted from the cognitive path atlas. The quantitative evaluation is completed by the following formula:

[0020]

[0021] where Score cog is the quantitative evaluation score; G ind is the individual cognitive path; G expert is the expert path; Dist graph is the graph edit distance function for calculating the difference between two graphs; NormFactor is the normalization factor for normalizing the evaluation score to a predetermined interval.

[0022] In an optional embodiment, the step of aggregating the individual cognitive paths to the cognitive path atlas and analyzing, specifically comprises:

[0023] superimposing the individual cognitive paths to the cognitive path atlas to form an updated cognitive path atlas reflecting the group behavior pattern;

[0024] analyzing the updated cognitive path atlas by graph mining algorithm to identify specific information representing the group behavior characteristics, the specific information including at least one of the following: expert path set, novice path set, common trap node, or stereotype path;

[0025] and outputting the identified specific information as the analysis result

[0026] In an optional embodiment, the analysis result is used to drive the step of generating a targeted virtual disturbance, specifically comprising: when a specific thought pattern path is contained in the analysis result, the cognitive confrontation engine generates a corresponding virtual disturbance, the function of which is to make the operation orientation following the thought pattern path fail in the state, and is used in the new training.

[0027] In another optional embodiment, the method further comprises, before the step of constructing the cognitive digital twin, generating a unique two-dimensional code identifier for the offline training equipment by an offline training base information management device, for identity recognition and management of the equipment.

[0028] In an optional embodiment, the method further comprises:

[0029] The terminal management device is used to visually display the individual cognitive path generated by the cognitive confrontation engine and the analysis result of the cognitive path atlas.

[0030] The second aspect of the application provides a secure training system based on a digital cloud platform, which comprises:

[0031] The offline training equipment is provided with:

[0032] A sensor array is used to collect the running parameters of the offline training equipment in real time.

[0033] A content synchronization interface;

[0034] A central processing device is internally provided with a cognitive confrontation engine, and is connected with the sensor array and the content synchronization interface, and is used to:

[0035] Construct a cognitive digital twin corresponding to the offline training equipment;

[0036] Receive the running parameters collected by the sensor array to synchronize to the cognitive digital twin in real time;

[0037] Receive the operation sequence of the student by the cognitive confrontation engine, and convert the operation sequence into an individual cognitive path; aggregate multiple individual cognitive paths to form a cognitive path atlas, and analyze the cognitive path atlas to generate an analysis result;

[0038] Generate a virtual disturbance based on the analysis result by the cognitive confrontation engine, inject the virtual disturbance into the cognitive digital twin, and drive the cognitive digital twin to perform state deduction;

[0039] Map the state change deduced by the cognitive digital twin to the offline training equipment via the content synchronization interface.

[0040] A terminal management device, connected with the central processing device, is used for:

[0041] Collecting the operation sequence of the trainee and sending it to the central processing device;

[0042] Receiving the individual cognitive path and the analysis result from the central processing device and performing visual display.

[0043] In summary, the present application includes at least one of the following beneficial technical effects:

[0044] 1. The present application converts the trainee's operation sequence into an individual cognitive path on a digital cloud platform, aggregates to form a group cognitive path atlas, and realizes dynamic self-adaptation of training. The system can analyze the atlas through graph mining algorithm to identify the group's mindset or common errors, and generate targeted virtual disturbances according to the analysis results. This feedback loop makes the safety training content no longer a static script, but can be personalized according to the real level and behavior pattern of the trainee, significantly improving the pertinence and effectiveness of the training.

[0045] 2. The present application models the trainee's operation process in safety training as a weighted directed graph carrying multi-dimensional information, and realizes objective quantitative evaluation of decision-making ability. This method not only records the operation sequence, but also quantifies cognitive factors such as decision-making reaction time and action complexity. By calculating the structural difference between the individual cognitive path graph of the trainee and the standard expert path graph on the cloud platform, the system can generate a stable and reproducible evaluation score. This overcomes the limitations of traditional safety training evaluation relying on the final result or subjective judgment, and provides a deeper, process-based evaluation basis.

[0046] 3. The present application provides a high-fidelity interactive safety training environment by constructing a cognitive digital twin that is in real-time linkage with offline training equipment. The digital twin deployed on the cloud platform not only receives real-time parameters collected by physical device sensors to maintain state synchronization, but also maps the state inference results generated by virtual disturbances in its internal to the physical device through the interface. This two-way, real-time information interaction between the digital world and the physical world enables trainees to train in real physical feedback, greatly enhancing the immersion and realism of safety training based on digital cloud platform. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The safety training system architecture based on digital cloud platform of the present application;

[0048] Figure 2 The safety training method flowchart based on digital cloud platform of the present application;

[0049] Figure 3 The adaptive training closed loop schematic diagram of the present application.

[0050] Wherein, 100, offline practical training equipment;110, sensor array;120, content synchronization interface;200, central processing device;210, cognitive confrontation engine;300, terminal management device;400, offline practical training base information management device;500, data statistics device. DETAILED DESCRIPTION

[0051] The following will be combined with the Figure 1 -Appendix Figure 3 The present application will be further described in detail.

[0052] The embodiment of the present application provides a kind of based on digital cloud platform security practical training method and system.

[0053] Refer to the Figure 1 , Figure 1 It is the security practical training system architecture diagram based on digital cloud platform according to an embodiment of the present application.The embodiment of the present application provides a kind of security practical training system deployed on digital cloud platform, and the system includes one or more offline practical training equipment 100, central processing device 200 and terminal management device 300.The central processing device 200 is connected with offline practical training equipment 100 and terminal management device 300.

[0054] The offline practical training equipment 100 is the entity device that student carries out physical operation.It is provided with sensor array 110, for real-time acquisition one or more operating parameters of the offline practical training equipment 100 in running process, for example temperature, pressure or speed, etc.Physical quantity.It is also provided with content synchronization interface 120 for offline practical training equipment 100, for receiving instruction from the central processing device 200, and changes one or more physical states of itself according to the instruction.

[0055] The central processing device 200 is the computing and control unit of the system, and it is deployed in cloud platform server.Cognitive confrontation engine 210 is integrated in the central processing device 200, and the engine is the processing module for executing core cognitive analysis and confrontation function of the present application.The central processing device 200 is connected with sensor array 110 and content synchronization interface 120 on the offline practical training equipment 100 through data interaction interface.

[0056] In one specific embodiment, the central processing device 200 can also be connected with an offline training base information management device 400. The offline training base information management device 400 can internally include an identification generation device and a map generation device. The identification generation device is used to generate a unique, machine-readable two-dimensional code identification for each offline training device 100 based on its unique identity information, such as device number or asset code, through a two-dimensional code generation algorithm. The map generation device is used to input the geographic position coordinate information of the base to generate a two-dimensional or three-dimensional virtual map including the position layout of all offline training devices 100. Students can start the training task for a specific device by scanning the two-dimensional code on the physical device through the terminal management device 300 or clicking the device annotation on the virtual map, which realizes the precise association between physical entities and digital training tasks.

[0057] In another specific embodiment, the central processing device 200 can also be connected with a data statistics device 500. The data statistics device 500 can include a multi-dimensional statistics device, an assessment evaluation module, and a visual display screen. The multi-dimensional statistics device collects and processes multi-source data generated during the training process through a database management system and data analysis algorithm, and calculates macro operation indicators such as total number of training participants, passing rate of each subject examination, and cumulative use time of each type of offline training device 100. The assessment evaluation module internally stores a preset assessment question bank and can generate a dedicated verification identifier according to instructions. Students trigger the assessment process by scanning the identifier and return the assessment results to the central processing device 200. The visual display screen displays the statistical results generated by the multi-dimensional statistics device and the device operation data stream collected by the sensor array 110 in real time in the form of a self-defined data board or a geographic distribution map, providing data support for training management.

[0058] The terminal management device 300 is a user terminal for students and management personnel to interact with the system, such as a personal computer, a tablet computer, or a dedicated console. It is in communication connection with the central processing device 200. The function of the terminal management device 300 is to collect the operation sequence generated by the students during the training and send the sequence to the central processing device 200 for processing. At the same time, the terminal management device 300 is also used to receive the processed results from the central processing device 200, such as formatted individual cognitive path data and group path analysis results, and perform visual display.

[0059] The system in this embodiment monitors and intervenes the state of offline practical training equipment 100 through the central processing device 200 set on the digital cloud platform, and collects the operation of students and feeds back the analysis results through the terminal management device 300. This overall architecture provides a basis for realizing dynamic adaptive personalized training and objective quantitative evaluation of the decision-making process of students.

[0060] Refer to the attached Figure 2 and Figure 3 , Figure 2 is a safety practical training method flowchart based on a digital cloud platform according to an embodiment of the present application. The method can include the following steps:

[0061] S100: Construct a cognitive digital twin corresponding to offline practical training equipment, and synchronize the initial operating parameters of the offline practical training equipment to the cognitive digital twin based on the sensor array set on the offline practical training equipment.

[0062] S200: The cognitive confrontation engine inside the central processing device generates a virtual disturbance and applies it to the cognitive digital twin. The cognitive digital twin deduces the state change caused by the disturbance and maps it to the offline practical training equipment, causing a physical state change. At the same time, the system records the operation sequence performed by the student to cope with the state change.

[0063] S300: Convert the operation sequence into a structured data model representing the decision-making process of the student, i.e., an individual cognitive path. The individual cognitive path models the student's operation, operation sequence, and related cognitive factors, providing a data basis for objective evaluation.

[0064] S400: Aggregate the individual cognitive path into a group cognitive path map and analyze the map to identify group behavior patterns. The analysis results are used to drive the cognitive confrontation engine to generate more targeted virtual disturbances in subsequent training, forming a self-adaptive training closed loop.

[0065] Refer to the attached Figure 1 and attached Figure 2 . Step S100 is to construct and synchronize the cognitive digital twin.

[0066] When a specific practical training task is started, the student uses the terminal management device 300 to scan the unique two-dimensional code identifier set on the offline practical training equipment 100. The two-dimensional code identifier is generated in advance by the offline training base information management device 400 based on the unique identity information of the offline practical training equipment 100. After completing the scanning identification, the terminal management device 300 sends the training start instruction containing the unique identity information of the offline practical training equipment 100 to the central processing device 200 deployed on the digital cloud platform through the communication network.

[0067] The central processing device 200 first parses the unique identity information of the offline practical training device 100 after receiving the training start instruction. According to the information, the central processing device 200 retrieves and calls the preset digital twin model corresponding to the offline practical training device 100 from the digital twin model library stored in the internal memory of the central processing device 200. The model is a structured data and program set that encapsulates the precise geometric dimensions, material properties, physical operation rules (such as fluid mechanics equations, Kirchhoff's law of electric circuit, etc.), and operation logic of all operable components of the offline practical training device 100. The central processing device 200 instantiates the model in the memory of the cloud server, thereby constructing the cognitive digital twin.

[0068] At the same time when the cognitive digital twin is completed, the central processing device 200 sends a real-time parameter collection instruction to the sensor array 110 arranged on the offline practical training device 100 through the data interaction interface. The sensor array 110 is not a single sensor, but a collection of multiple types of sensors, which are arranged on the key measurement nodes of the device according to the specific physical characteristics of the offline practical training device 100. For example, if the offline practical training device 100 is a chemical pipeline system, the sensor array 110 can include temperature sensors installed on the pipeline wall, pressure sensors installed on the pressure vessel, speed sensors installed on the pump, and opening position sensors installed on the valve.

[0069] Each sensor in the sensor array 110 immediately collects the precise values of various operating parameters of the offline practical training device 100 at the current time after receiving the collection instruction. These collected operating parameter values (for example, temperature is 25.3°C, pressure is 1.05 MPa, pump speed is 0 RPM, and No. 2 valve opening is 0%) are packaged into an initial state data set and transmitted to the central processing device 200 by the sensor array 110 through wired or wireless means.

[0070] After the central processing device 200 receives the data set, it writes the parameter values in the data set into the corresponding data fields in the cognitive digital twin. This process completes the synchronization of the initial state of the cognitive digital twin, ensuring that the model state in the digital world and the entity device state in the physical world are completely consistent at the zero time of the start of the training.

[0071] Referring to the accompanying drawings Figure 1 - the accompanying drawings Figure 3 The step S200 of the method in the embodiment of the application is described in detail. This step is to implement cognitive confrontation and state mapping.

[0072] After the cognitive digital twin is completed and the state is synchronized, the cognitive confrontation engine 210 deployed in the central processing device 200 is used to generate a virtual disturbance. The generation of the virtual disturbance can be based on one of two modes:

[0073] One is to set a predetermined training strategy, for example, to set a basic fault scene for a primary student;

[0074] The second is to analyze the cognitive path map generated in the subsequent step S400, for example, to specially design a scene to break the identified group thinking mode.

[0075] The virtual disturbance is manifested as an abnormal state value of one or a group of specific parameters at the data level, such as a short circuit of an analog circuit or a signal simulating a valve stuck. The cognitive confrontation engine 210 applies the virtual disturbance to the instantiated cognitive digital twin. After receiving the disturbance, the physical rule engine (for example, a circuit simulation model or a fluid dynamics solver) encapsulated in the cognitive digital twin immediately performs state deduction, calculates a series of continuous state changes caused by the disturbance and evolving over time according to the disturbance and the current device state.

[0076] For example, if the virtual disturbance is an abnormal drop in the power supply voltage of pump No. 2, the state changes deduced by the cognitive digital twin can include a series of quantitative data of chain reactions, such as a decrease in the pump speed, a decrease in the fluid pressure in the corresponding pipeline, and an increase in the temperature of the downstream device due to the slow flow rate.

[0077] The central processing device 200 sends this set of deduced state change data to the content synchronization interface 120 on the offline training device 100 through the communication network. The function of the content synchronization interface 120 is to convert these digital state variables into control instructions for specific actuators or displays on the offline training device 100. For example, the data of reducing the pressure to 0.8 MPa is converted into an instruction to control the pointer of the pressure gauge to deflect to the 0.8 MPa scale; the reduction of the speed to 500 RPM is converted into an instruction to control the actual pump speed to decrease through the frequency converter. This process realizes the state mapping from the digital model to the physical entity, so that the student can perceive the physical state change consistent with the real fault on the offline training device 100.

[0078] At the same time, the system records the series of physical operations performed by the student to respond to this state change through the terminal management device 300 or directly through the sensors arranged on the device operation panel, forming an operation sequence. The operation sequence is an ordered event set including a time stamp, which accurately records each action performed by the student, the time when the action occurs, and the object of the action, providing original input data for the decision process analysis in the subsequent step S300.

[0079] Referring to the drawings Figure 1 -attached Figure 3 The operation sequence recorded in step S200 is sent to the central processing device 200, and step S300 is performed by the cognitive confrontation engine 210, i.e. transforming the operation sequence into a structured data model representing the decision-making process of the trainee, i.e. an individual cognitive path.

[0080] The operation sequence is a time-ordered event tuple sequence, the structure of which can be represented as In the formula, a i is the ith action performed by the trainee, such as pressing the start button or rotating the valve; t i is the timestamp of the action; o i is the object of the action, such as pump No. 1 or valve No. 3; k is the total number of operations performed by the trainee in one training; and i is the sequence number of the operation.

[0081] The cognitive confrontation engine 210 transforms the event tuple sequence into a weighted directed graph G ind =(V,E,W) as the individual cognitive path.

[0082] In this transformation process, each event tuple (a i ,t i ,o i ) in the sequence is mapped to a node v i in the graph, and all nodes form a node set V. For any two event tuples (a i-1 ,t i-1 ,o i-1 ) and (a i ,t i ,o i ) in the sequence that are consecutive in time, a directed edge e i-1 from node v i to node v i is created in the graph, and all directed edges form an edge set E °

[0083] The weight set W in the weighted directed graph is composed of a weight vector w i associated with each edge e i . The weight vector w i quantifies the cognitive process of the trainee between two consecutive operations and can be calculated by the following formula:

[0084] w i =[Δt i ,C(a i ),F(o i-1 )];

[0085] wherein, At i = t i - t i-1 , denotes the decision reaction time; C(a i ) is the quantified value of the complexity of action a i ; F(o i-1 ) is the state feature feedback of the object o i-1 before the action is performed.

[0086] In an optional embodiment, the step can also quantitatively evaluate the decision-making process of the trainee. The cognitive confrontation engine 210 extracts one or more standard expert paths G expert from the cognitive path atlas, and calculates the difference between the individual cognitive path Dist graph of the trainee and the standard expert paths, to generate a quantitative evaluation score, according to the following formula:

[0087]

[0088] wherein, Score cog is the quantitative evaluation score; G ind is the individual cognitive path; G expert is the expert path; Dist graph is a graph edit distance function used to calculate the difference between two graphs, and the result represents the minimum cost required to convert the trainee path into the expert path; NormFactor is a normalization factor set to normalize the score to a specific interval (e.g. [0, 1]). This step overcomes the subjectivity and one-sidedness of traditional evaluation methods by modeling the operation process as a weighted directed graph carrying multi-dimensional information, and realizes objective and reproducible quantitative evaluation of decision-making ability.

[0089] Refer to Appendix 1-Appendix Figure 3 Step S300 is aggregation analysis and feedback formation.

[0090] The individual cognitive path G ind of the trainee generated in step S300 is submitted to the central processing device 200. The central processing device 200 superimposes and aggregates the individual cognitive path with a cognitive path atlas composed of historical data stored in the digital cloud platform. The aggregation operation is technically realized by performing the union operation of graphs, i.e. merging the nodes and directed edges in the individual cognitive path into the cognitive path atlas. For the same nodes (i.e. nodes representing the same operation and object entity) that already exist in the atlas, the occurrence frequency and other statistical attributes are increased.

[0091] The central processing unit 200 is used to perform graph mining algorithms on the cognitive path graph after it has been updated, in order to identify specific information characterizing group behavior. In one specific embodiment, the graph mining algorithm used may include: a frequent subgraph mining algorithm for discovering frequently occurring operation sequences, and a centrality analysis algorithm for identifying key nodes.

[0092] The frequent subgraph mining algorithm identifies mental path patterns by calculating the support of subgraphs. Subgraph g is located in the cognitive path graph database D (which contains multiple individual cognitive paths G). i The support level supp(g) in the data can be calculated using the following formula:

[0093] In the formula, Subgraph g represents the individual cognitive path graph G. i A substructure; |·| represents the number of elements in the set. When the support supp(g) of a subgraph g is greater than a preset minimum support threshold min_supp, the subgraph g is identified as a frequent subgraph, i.e., a mindset path.

[0094] The centrality analysis algorithm can employ the betweenness centrality algorithm to identify common trap nodes. The betweenness centrality C of a node v is... B (v) can be calculated using the following formula:

[0095]

[0096] In the formula, s and t are any two nodes in the graph; σ st σ is the total number of all shortest paths from node s to node t; st (v) is the number of nodes v that are passed through in these shortest paths.

[0097] The identified specific information includes at least one of the following:

[0098] Expert path set: Scores evaluated from multiple historical records cog Individual cognitive paths exceeding a preset threshold constitute an efficient and correct operational process.

[0099] Novice path set: consisting of multiple assessment scores below a preset threshold, or decision-making reaction time Δt i The significantly longer individual cognitive pathways reflect the common behavioral patterns of novice learners.

[0100] Common trap node: a node in the cognitive path map that is connected to multiple low-score or failure state paths. This node represents a device state or operation node that a trainee commonly makes a wrong judgment.

[0101] Mindset path: a frequently executed but not optimal fixed operation sequence discovered by the frequent subgraph mining algorithm.

[0102] The analysis result is fed back to the cognitive adversarial engine 210 as input for dynamically generating more targeted virtual perturbations in subsequent new training. For example, when a specific mindset path is identified in the analysis result, the cognitive adversarial engine 210 generates a corresponding targeted new virtual perturbation. The role of this virtual perturbation is that if the trainee still follows the identified mindset path for operation, it will lead to a preset training failure state. This design forces the trainee to explore new operation strategies to cope with the perturbation, thereby breaking the fixed mindset.

[0103] The feedback loop makes the safety training content no longer static scripts, but can be dynamically adjusted and evolved according to the real level and behavior patterns of the trainee group, realizing the self-adaptation of the training. In addition, the analysis result, such as the common trap node and the mindset path highlighted in the form of graphics, can be sent to the terminal management device 300 for visualization display, providing data support for trainee review and instructor guidance.

[0104] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements 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 digital cloud platform-based security practical training method, characterized in that, The method comprises the following steps: constructing a cognitive digital twin corresponding to the offline training equipment by a central processing device, and synchronizing the running parameters of the offline training equipment to the cognitive digital twin in real time based on a sensor array on the offline training equipment; after querying a cognitive path atlas based on a cognitive adversarial engine in the central processing device, dynamically generating a virtual disturbance and injecting it into the cognitive digital twin, the cognitive digital twin performing state deduction according to preset physical rules, and mapping the deduced state changes to the offline training equipment via a content synchronization interface, while the cognitive adversarial engine records the operation sequence of the trainee for the offline training equipment; the cognitive adversarial engine converts the operation sequence into an individual cognitive path representing the decision-making process of the trainee; aggregating the individual cognitive path into the cognitive path atlas, analyzing the cognitive path atlas, and outputting the analysis result, which is used to drive the cognitive adversarial engine to generate targeted virtual disturbances in a new training; the step of aggregating the individual cognitive path into the cognitive path atlas and analyzing the cognitive path atlas and outputting the analysis result comprises: superimposed aggregation of the individual cognitive path into the cognitive path atlas to form an updated cognitive path atlas reflecting group behavior patterns; analyzing the updated cognitive path atlas by a graph mining algorithm to identify specific information representing group behavior characteristics, the specific information including at least one of the following: an expert path set, a novice path set, a common trap node, or a mindset path; and outputting the identified specific information representing group behavior characteristics as the analysis result; the analysis result is used to drive the cognitive adversarial engine to generate targeted virtual disturbances in a new training, specifically including: when the analysis result includes a specific mindset path, the cognitive adversarial engine generates a corresponding virtual disturbance, which causes the operation following the mindset path to fail and is used in a new training.

2. The digital cloud platform-based security practical training method according to claim 1, characterized in that, the step of converting the operation sequence into an individual cognitive path representing the decision-making process of the trainee by the cognitive adversarial engine comprises: the cognitive adversarial engine converts the operation sequence into a weighted directed graph as the individual cognitive path; The sequence of operations is a time-ordered sequence of event tuples wherein a i is the ith action performed by the learner; t i is the timestamp at which the action occurred; o i is the object on which the action acted; and k is the total number of operations performed by the learner in one training. The form of the weighted directed graph as an individual cognitive path is G ind = (V, E, W); where V is a node set, E is an edge set, and W is a weight set; Each node v in the set of nodes V i Corresponding to one event (a i ,t i ,o i ) in the sequence of operations, each edge e i = (v i-1 ,v i ) in the set of edges E represents a sequential flow between events and is associated with one weight vector W i from the set of weights W.

3. The digital cloud platform-based security practical training method according to claim 2, characterized in that, The weight vector w i The calculation formula is: w i = [At i , C(a i ), F(o i-1 )]; where Δt i = t i -t i-1 , denotes the decision reaction time; C(a i ) is a quantized value of the complexity of action a i ; F(o i-1 ) is the state feature feedback of the object o i-1 on which the action acts before the action is performed.

4. The digital cloud platform-based security practical training method according to claim 1, characterized in that, after the step of converting the operation sequence into an individual cognitive path representing the decision-making process of the trainee by the cognitive adversarial engine, further comprising: the cognitive adversarial engine quantitatively evaluates the decision-making process of the trainee by graph matching calculation of the individual cognitive path with an expert path extracted from the cognitive path atlas.

5. The digital cloud platform-based security practical training method according to claim 4, characterized in that, the quantitative evaluation is completed by the following formula: where Score cog is the quantitative assessment score; G ind is the individual cognitive path; G expert is the expert path; Dist graph is a graph edit distance function for computing the difference between two graphs; NormFactor is a normalization factor.

6. The digital cloud platform-based security practical training method according to claim 1, characterized in that, before the step of constructing a cognitive digital twin corresponding to the offline training equipment by a central processing device, further comprising: generating a unique two-dimensional code identifier for the offline training equipment by an offline training base information management device.

7. The digital cloud platform-based security practical training method according to claim 1, characterized in that, The cognitive confrontation engine converts the generated individual cognitive path and the analysis result of the cognitive path atlas, and visualizes the same through a terminal management device.

8. A digital cloud platform-based security practical training system applied to the digital cloud platform-based security practical training method according to any one of claims 1-7. Comprise: Offline training equipment, which is provided with: A sensor array for collecting the operating parameters of the offline training equipment in real time; A content synchronization interface; A central processing device internally provided with a cognitive confrontation engine, which is connected with the sensor array and the content synchronization interface, and is used for: Building a cognitive digital twin corresponding to the offline training equipment; Receiving the operating parameters collected by the sensor array to synchronize to the cognitive digital twin in real time; Through the cognitive confrontation engine, receiving the operation sequence of the student, and converting the operation sequence into an individual cognitive path; Aggregating a plurality of individual cognitive paths to form a cognitive path atlas, and analyzing the cognitive path atlas to generate an analysis result; Through the cognitive confrontation engine, dynamically generating a virtual disturbance based on the analysis result, injecting the virtual disturbance into the cognitive digital twin, and driving the cognitive digital twin to perform state deduction; Mapping the state changes deduced by the cognitive digital twin to the offline training equipment through the content synchronization interface; A terminal management device connected with the central processing device, and used for: Collecting the operation sequence of the student and sending it to the central processing device; Receiving the individual cognitive path and the analysis result from the central processing device and visualizing the same.

Images