A building construction experience method and system based on VR technology

Abstract

The application discloses a kind of based on VR technology's building construction experience method and system, it is related to construction safety training technical field, method includes: generating building construction basic scene model based on digital twin technology, complexity is adjusted by machine learning algorithm and implants dynamic risk hidden danger, forms dynamic virtual scene;Multiple risk trigger thresholds are set based on the scene, generate trigger result and build matched emergency scene sequence;Call emergency scene sequence, collect user operation behavior data and physiological characteristic signal;Synchronous operation data, guide mark potential risk point, analyze data correlation and identify dangerous operation mode;Record three-dimensional motion trajectory, generate holographic evaluation report;Based on report, build user cognitive defect atlas, analyze associated emergency scene sequence in virtual component safety parameter, carry out targeted re-experience training.The application solves the problem that traditional training is abstract, existing VR training only corrects operation without repairing cognition, and improves the pertinence and effectiveness of training.

Description

Technical Field

[0001] This invention relates to the field of construction safety training technology, and in particular to a construction construction experience method and system based on VR technology. Background Technology

[0002] Current construction safety training relies heavily on classroom lectures, video viewing, or on-site observation, with less than ideal results. Classroom and video teaching tend to be abstract, making it difficult for trainees to truly experience dangerous construction scenarios, resulting in weak understanding and retention of safety knowledge. While on-site observation is more intuitive, construction sites themselves are high-risk and are limited by time constraints and space, making large-scale implementation difficult and costly.

[0003] Existing technologies attempt to use VR to construct virtual construction scenarios for training, which to some extent solves the safety and scenario limitations of traditional training. However, most current VR construction training only focuses on correcting operational behaviors and fails to address trainees' cognitive blind spots. The system simply points out operational errors but cannot analyze the underlying causes of potential hazards, leading to repeated errors by trainees and significantly reducing the training effectiveness. Summary of the Invention

[0004] To address the technical problem that most existing VR construction training technologies only focus on correcting operational behaviors and fail to address the cognitive blind spots of trainees, this invention provides a VR-based construction experience method and system.

[0005] The technical solution adopted in this invention is:

[0006] The first aspect of this application provides a method for experiencing building construction based on VR technology, including the following steps:

[0007] Step 1: Generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model and embed dynamic risks and hidden dangers through machine learning algorithms to form a dynamic virtual scene;

[0008] Step 2: Set multi-level risk trigger thresholds based on dynamic virtual scenarios. The multi-level risk trigger thresholds correspond to different danger levels in the scenario. When the state parameters of dynamic risk hazards reach the corresponding level of multi-level risk trigger thresholds, a trigger result is generated. An emergency scenario sequence matching the trigger results at each level is constructed using a branch logic structure.

[0009] Step 3: Based on the trigger result, invoke the corresponding emergency scenario sequence, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through the sensing and interaction device;

[0010] Step 4: Synchronize the operational behavior data to the corresponding emergency scenario sequence, guide the user to mark potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operational behavior data and physiological characteristic signals based on the potential risk point marking results to identify dangerous operation modes.

[0011] Step 5: Record the user's three-dimensional motion trajectory in the emergency scenario sequence, and generate a holographic assessment report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes;

[0012] Step 6: Construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, use the scenario backtracking tool to analyze the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in Step 2. Conduct targeted re-experience training in this emergency scenario sequence.

[0013] Preferred: Step 1 includes the following sub-steps:

[0014] Step 1.1: Collect 3D point cloud data, building information model and historical accident case data of each stage of building construction, and fuse the above data based on digital twin technology to generate a basic scene model of building construction.

[0015] Step 1.2, determine whether there are user historical operation records: if there are, use machine learning algorithms to parse the user historical operation records, establish a matching model between scene complexity and user operation ability, and adjust the scene complexity of the building construction basic scene model based on the matching model; if there are no such records, use preset scene complexity parameters to adjust the scene complexity of the building construction basic scene model.

[0016] Step 1.3: Incorporate dynamic risks and hazards into the adjusted building construction basic scenario model. Dynamic risks and hazards are potential safety risks that change state with time parameters or user operation intervention.

[0017] Step 1.4: Integrate and adjust the basic construction scenario model and dynamic risk hazards after adjusting the scenario complexity to form a dynamic virtual scenario.

[0018] Preferably, step 2 includes the following sub-steps:

[0019] Step 2.1: Based on the scene parameters of the dynamic virtual scene, set multi-level risk trigger thresholds. The multi-level risk trigger thresholds include the early warning threshold for the initial state of the corresponding hidden danger, the intervention threshold for the evolution state of the corresponding hidden danger, and the accident threshold for the failure state of the corresponding hidden danger.

[0020] Step 2.2: Monitor the status parameters of dynamic risks and hazards in the dynamic virtual scene in real time. When the status parameters of dynamic risks and hazards reach any level of the multi-level risk trigger threshold, generate a trigger result containing the threshold level and hazard characteristic information.

[0021] Step 2.3: Using a branch logic structure, based on the user's operational response efficiency and operational accuracy in the dynamic virtual scene, construct emergency scenario subsequences that match the trigger results at each level. Different emergency scenario subsequences contain differentiated hazard evolution paths.

[0022] Step 2.4: Integrate the emergency scenario subsequences corresponding to the triggering results at each level to form an emergency scenario sequence.

[0023] Preferably, step 3 includes the following sub-steps:

[0024] Step 3.1: Receive the trigger result generated in Step 2, and call the corresponding level of the scenario in the emergency scenario sequence constructed in Step 2 according to the threshold level in the trigger result;

[0025] Step 3.2: Collect user operation behavior data in the corresponding level of emergency scenario through sensor interaction device. The operation behavior data includes hand operation trajectory, limb movement angle and device operation force.

[0026] Step 3.3: Collect physiological characteristic signals of users in the corresponding level of emergency scenario through sensor interaction device. The physiological characteristic signals include heart rate variability, skin conductance response and electromyographic activity intensity. The collection frequency of physiological characteristic signals is synchronized with the frame rate of dynamic virtual scene.

[0027] Step 3.4: Associate the operational behavior data and physiological characteristic signals with the timeline of the corresponding emergency scenario.

[0028] Preferably, step 4 includes the following sub-steps:

[0029] Step 4.1: Establish a mapping relationship between physical operation parameters and virtual actions in the dynamic virtual scene, and based on this mapping relationship, synchronize the operation behavior data collected in Step 3 to the corresponding level of emergency scene;

[0030] Step 4.2 guides users to mark potential risk points in the corresponding level of emergency scenarios. The marking of potential risk points adopts a combination of three-dimensional coordinate annotation and semantic tags. The semantic tags include the type of hazard, risk level and scope of impact.

[0031] Step 4.3: Use time-series comparison method to analyze the correlation between operational behavior data and physiological characteristic signals, and identify the causal relationship between abnormal operational behavior and fluctuations in physiological characteristic signals;

[0032] Step 4.4: Based on the correlation analysis results and the completeness and accuracy of the potential risk point markings, identify hazardous operation modes, including path deviation, response delay, and handling error.

[0033] Preferably, step 5 includes the following sub-steps:

[0034] Step 5.1: Record the user's three-dimensional movement trajectory in the corresponding level of emergency scenario using spatial positioning equipment;

[0035] Step 5.2: Extract the operational behavior data and physiological characteristic signals collected in Step 3, the potential risk point marking results and dangerous operation patterns generated in Step 4, and construct a multi-dimensional analysis dataset;

[0036] Step 5.3: Quantify the multi-dimensional analysis dataset to generate quantitative results for operational compliance score, risk identification accuracy and physiological stress index;

[0037] Step 5.4: Integrate the three-dimensional motion trajectory, quantitative processing results, and dangerous operation modes to generate a holographic assessment report containing visual analysis content.

[0038] Preferably, step 6 includes the following sub-steps:

[0039] Step 6.1: Based on the quantitative processing results and dangerous operation modes in the holographic assessment report, construct a user cognitive deficiency map. The user cognitive deficiency map includes cognitive shortcomings in risk identification, handling logic, and emergency response dimensions. The degree of cognitive shortcomings in each dimension is quantified by weighting.

[0040] Step 6.2: Based on the cognitive shortcomings with higher weights in the user cognitive deficiency map, locate the multi-level risk trigger threshold level and emergency scenario sequence corresponding to Step 2;

[0041] Step 6.3: Freeze the keyframes of the target emergency scenario sequence using the scenario backtracking tool, and analyze the safety parameters of the virtual components in the target emergency scenario sequence. The safety parameters of the virtual components include the structural bearing limit, equipment operating threshold, and effective range of protective facilities.

[0042] Step 6.4: Reproduce the associated hidden danger scenarios in the emergency scenario sequence of the corresponding level, conduct targeted re-experience training, and verify the effect of correcting cognitive shortcomings through the training process.

[0043] The second aspect of this application provides a VR-based construction experience system, and the above-mentioned VR-based construction experience method includes:

[0044] A dynamic virtual scene construction module is used to generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model through machine learning algorithms and embed dynamic risks and hidden dangers to form a dynamic virtual scene.

[0045] The risk threshold setting and emergency scenario construction module is used to set multi-level risk trigger thresholds based on dynamic virtual scenarios. The multi-level risk trigger thresholds correspond to different danger levels in the dynamic virtual scenarios. When the state parameters of dynamic risk hazards reach the corresponding level of multi-level risk trigger thresholds, a trigger result is generated. An emergency scenario sequence matching the trigger results at each level is constructed using a branch logic structure.

[0046] A multimodal data acquisition module is used to call the corresponding emergency scenario sequence according to the trigger result, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through a sensing and interaction device;

[0047] The operation synchronization and hazard pattern recognition module is used to synchronize operation behavior data to the corresponding emergency scenario sequence, guide the user to complete the marking of potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operation behavior data and physiological characteristic signals by combining the potential risk point marking results to identify dangerous operation patterns.

[0048] The three-dimensional trajectory recording and holographic evaluation module is used to record the user's three-dimensional motion trajectory in an emergency scenario sequence, and generate a holographic evaluation report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes.

[0049] The cognitive deficiency analysis and targeted training module is used to construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, the module analyzes the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in the risk threshold setting and emergency scenario construction module through the scenario backtracking tool, and conducts targeted re-experience training in the emergency scenario sequence.

[0050] The beneficial effects of this invention are: the user cognitive deficiency map constructed based on the holographic assessment report can clearly identify the shortcomings of trainees in risk cognition; the analysis of the safety parameters of virtual components through the scenario retrospective tool can help trainees understand the underlying causes of risk hazards, rather than just knowing the operational errors; the targeted re-experience training for cognitive shortcomings can focus on the trainees' weak cognitive links, avoid blindly repeating training, and improve the pertinence and efficiency of training. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the method of Embodiment 1 of the present invention;

[0052] Figure 2 This is a structural block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0053] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0054] Example 1 provides a construction experience method based on VR technology, such as Figure 1 As shown, it includes the following steps:

[0055] Step 1: Generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model through machine learning algorithms and embed dynamic risks and hidden dangers to form a dynamic virtual scene.

[0056] It should be noted that the basic construction scene model refers to a 3D digital model based on the actual construction scene, including the construction site layout (such as material storage areas and construction access roads), building structural elements (such as walls, beams, columns, and scaffolding), and construction equipment models (such as tower cranes and concrete pump trucks). It serves as the basic carrier of the dynamic virtual scene. Digital twin technology refers to collecting construction scene data from the physical world to construct a digital model that maps to the physical scene in a 1:1 ratio, and synchronizing the state of the physical scene with the digital model to ensure the authenticity and timeliness of the digital model.

[0057] Machine learning algorithms refer to supervised learning algorithms (such as random forest algorithms) used in this step for evaluating user operational capabilities. By extracting features from users' historical operation data and training the model, a mapping relationship between user operational capabilities and scene complexity is established.

[0058] Scene complexity refers to the set of parameters in the basic construction scene model that affect the difficulty of user operation, including environmental interference factors (such as the digital simulation intensity of pedestrian flow and equipment operating noise at the construction site), task complexity (such as single-task and cross-task types), and risk point density (such as the number of potential safety hazards in a unit area). Dynamic risks and hazards refer to potential safety risks whose state changes with time parameters (such as the progress of construction) or user operation intervention (such as unauthorized adjustment of scaffolding pole spacing). Their state parameters (such as the scope of impact and degree of danger) evolve dynamically with the triggering conditions, which is different from the fixed risk points in static scenes.

[0059] Dynamic virtual scenes refer to interactive 3D virtual environments that integrate and adjust the complexity of basic construction scene models with dynamic risks and hazards, and have the characteristics of real-time updates of scene status and real-time response to user operations.

[0060] The core objective of this step is to construct an interactive virtual platform that closely resembles the real construction environment. On the one hand, digital twin technology ensures the authenticity of the scenario, solving the problem of abstraction in traditional classroom training. On the other hand, machine learning algorithms dynamically adjust the complexity to achieve personalized instruction, avoiding the problem of novice users giving up due to overly difficult scenarios and experienced users not improving due to overly simplistic scenarios. At the same time, dynamic risks and hidden dangers are embedded to simulate the evolution of risks from their inception to their outbreak in real construction, solving the problem of static risks and disconnect from reality in existing VR scenarios.

[0061] In one possible implementation: step 1 includes the following sub-steps:

[0062] Step 1.1: Collect 3D point cloud data, building information model and historical accident case data of each stage of building construction, and fuse the above data based on digital twin technology to generate a basic scene model of building construction.

[0063] For example, 3D point cloud data can be obtained by scanning the construction site and the main structure of the building with a laser scanner to obtain spatial point clouds with millimeter-level precision, which include the geometric shape and spatial location information of the objects; Building Information Modeling (BIM) can import BIM models from the design phase to extract the structural parameters of the building (such as the cross-sectional dimensions of beams and the concrete strength grade) and the technical parameters of the construction equipment (such as the maximum lifting capacity and operating radius of the tower crane); Historical accident case data can collect safety accident reports of similar construction scenarios in the past 5 years to extract the scene characteristics of the accident (such as the accident location, the equipment involved, and the triggering conditions).

[0064] Based on the multi-source data fusion framework of digital twin technology, the three-dimensional point cloud data is registered with the BIM model (the spatial coordinate alignment is achieved by using the Iterative Closest Point (ICP) algorithm), so that the structural parameters of the BIM model and the geometry of the point cloud data are accurately matched. Then, the scene features in the historical accident case data are mapped to the fused model (such as marking the location of the "scaffolding collapse accident" as a key area in the model), generating a basic scene model of building construction.

[0065] Step 1.2, determine whether there are user historical operation records: if they exist, parse the user historical operation records using machine learning algorithms, establish a matching model between scene complexity and user operation ability, and adjust the scene complexity of the building construction basic scene model based on the matching model; if they do not exist, adjust the scene complexity of the building construction basic scene model using preset scene complexity parameters.

[0066] For example, firstly, the system queries the user's identity-related operation record database to determine if the user has historical operation records. If historical operation records exist, feature parameters (including operation completion time, number of operation errors, and risk identification accuracy) are extracted from the historical operation records. The feature parameters are then trained using a random forest algorithm. Users with operation completion time ≤ 80% of the industry average, number of errors ≤ 2, and risk identification accuracy ≥ 80% are labeled as having higher operational ability, while those with the opposite are labeled as having lower operational ability. A user operational ability classification model (i.e., a matching model between scenario complexity and user operational ability) is generated through training. The model outputs three levels: low complexity, medium complexity, and high complexity.

[0067] If no historical operation records exist, a preset scenario complexity parameter is used. This parameter is determined based on industry standards or internal training annotations. The default initial complexity is low, and it will be dynamically updated using machine learning algorithms after subsequent user operation data is collected. The parameters of the basic construction scenario model are adjusted according to the matching model output or the preset complexity level.

[0068] Low complexity adjustment refers to reducing environmental interference factors (shielding the pedestrian flow simulation model in non-core construction areas, reducing the acoustic simulation intensity of equipment operating noise), simplifying work tasks (retaining only a single work scenario, such as an independent scaffolding erection task, without other cross-operations), and reducing the density of risk points (retaining only 2-3 core risk points in a unit area, such as excessive spacing between scaffolding uprights).

[0069] High complexity adjustment refers to increasing environmental interference factors (enabling full-scene pedestrian flow simulation, including the dynamic paths of construction personnel and material transport vehicles), increasing the complexity of work tasks (setting up cross-operation scenarios, such as tower crane hoisting operations and scaffolding erection operations being carried out simultaneously), and increasing the density of risk points (setting 5-6 risk points per unit area, including excessive pole spacing, incomplete scaffolding board laying, and damaged safety nets).

[0070] Step 1.3: Incorporate dynamic risk hazards into the adjusted building construction basic scenario model. Dynamic risk hazards are potential safety risks whose state changes with time parameters or user operation intervention.

[0071] For example, based on industry standards such as the "Technical Specification for Safety of High-Altitude Operations in Building Construction" and the "Safety Regulations for Scaffolding", the types of dynamic risks that frequently occur in building construction are screened out, including but not limited to: displacement of temporary support structures (such as lateral displacement of scaffolding uprights due to uneven stress); overload of electrical equipment (such as temporary distribution boxes exceeding the rated current due to too many devices connected); and failure of fall protection at heights (such as safety rope hooks falling off due to vibration).

[0072] For each type of dynamic risk hazard, trigger conditions and parameter change rules for state evolution are defined. Taking scaffolding upright displacement as an example: the trigger condition is that the user fails to install the ground bracing according to specifications; the evolution rule is that the displacement of the scaffolding upright increases linearly with time, and the force parameters (such as bending moment and shear force) of the upright in the model are updated synchronously with each displacement change. The evolution rules of dynamic risks hazard are written into the logic layer of the building construction foundation scenario model, so that risks hazard can automatically trigger state changes based on user operations or time parameters, and the state parameters of the risk (displacement, current value, hook firmness) are written into the scenario database in real time for subsequent steps to call.

[0073] Step 1.4: Integrate and adjust the basic construction scenario model and dynamic risk hazards after adjusting the scenario complexity to form a dynamic virtual scenario.

[0074] For example, the scene integration engine integrates the basic construction scene model with adjusted complexity from step 1.2 with the logic layer embedding dynamic risks and hazards from step 1.3. This ensures that changes in the state of risks and hazards are reflected in the scene's visual presentation (e.g., users can visually see the pole tilting when it shifts) and interactive responses (e.g., the indicator light on the distribution box changes from green to red when electrical equipment is overloaded). By simulating user operations (e.g., triggering the risk of scaffold pole displacement), the smoothness of risk evolution and the effectiveness of complexity parameters in the scene are verified. After ensuring that the dynamic virtual scene has no logical loopholes, the result is output to step 2.

[0075] Step 2: Set multi-level risk trigger thresholds based on the dynamic virtual scene. The multi-level risk trigger thresholds correspond to different danger levels in the dynamic virtual scene. When the state parameters of the dynamic risk hazard reach the corresponding level of the multi-level risk trigger threshold, a trigger result is generated. An emergency scene sequence matching the trigger results at each level is constructed using a branch logic structure.

[0076] It should be noted that multi-level risk trigger thresholds refer to the critical values ​​set for the state parameters of dynamic risks and hazards to distinguish different levels of danger. These include three levels: warning threshold, intervention threshold, and accident threshold. Each level corresponds to a specific state of the dynamic risk and hazard, serving as the basis for judging the stage of risk evolution. The state parameters of dynamic risks and hazard refer to quantitative indicators describing the current state of the dynamic risk and hazard. For example, the state parameter for pole displacement risk is the displacement amount, and the state parameter for electrical equipment overload risk is the actual current value. These values ​​change in real time as the risk evolves. The trigger result refers to the structured data generated by the system when the state parameters of a dynamic risk and hazard reach a certain threshold level. This data includes the threshold level, hazard type, current state parameter value, and trigger time, and is used to drive the subsequent invocation of emergency scenarios.

[0077] The branching logic structure refers to a scenario transition logic framework built based on user operation selections and trigger results. Different operation selections correspond to different scenario branches, ensuring the relevance between emergency scenarios and user operations and avoiding a disconnect between scenarios and operations. An emergency scenario sequence refers to a set of scenarios consisting of multiple emergency scenario sub-sequences, covering different stages of risk evolution. Each emergency scenario sub-sequence corresponds to a combination of a certain level of trigger result and a specific user operation, including risk handling tasks, scenario feedback logic, and other content.

[0078] This step addresses the issue of ambiguous risk levels by setting multi-level thresholds to divide the evolution of dynamic risks into quantifiable stages; it also ensures timely scenario feedback by automatically linking risk status and scenario response through trigger results; and it constructs emergency scenario sequences through branching logic, allowing users to experience different risk handling processes under different operational choices, thus solving the problem of scenario monotony in existing VR training and providing diverse scenario carriers for subsequent data collection.

[0079] In one possible implementation, step 2 includes the following sub-steps:

[0080] Step 2.1: Based on the scene parameters of the dynamic virtual scene, set multi-level risk trigger thresholds. The multi-level risk trigger thresholds include the early warning threshold for the initial state of the corresponding hidden danger, the intervention threshold for the evolution state of the corresponding hidden danger, and the accident threshold for the failure state of the corresponding hidden danger.

[0081] For example, for each dynamic risk hazard implanted in step 1.3, its core state parameters (such as the displacement amount of the pole displacement risk and the actual current value of the electrical overload risk) are extracted, and the value range of each level of threshold is determined by referring to the risk threshold value in the "Construction Safety Inspection Standard".

[0082] The warning threshold for the initial state of a potential hazard corresponds to the risk budding stage. When the state parameters reach this value, the risk has not yet posed a direct threat to construction safety, but trainees should be alerted. The intervention threshold for the evolution state of a potential hazard corresponds to the risk development stage. When the state parameters reach this value, the risk has posed a potential threat to construction safety, and trainees need to take appropriate measures. The accident threshold for the failure state of a potential hazard corresponds to the risk outbreak stage. When the state parameters reach this value, the risk has led to a safety accident, and trainees need to carry out emergency response procedures.

[0083] Each threshold level is bound to its corresponding dynamic risk hazard type and stored in a threshold database. A real-time communication interface is established between the thresholds and the dynamic virtual scenario to ensure that changes in risk status parameters in the scenario can be compared with the thresholds in real time.

[0084] Step 2.2: Monitor the status parameters of dynamic risks and hazards in the dynamic virtual scene in real time. When the status parameters of dynamic risks and hazards reach any level of the multi-level risk trigger threshold, generate a trigger result containing the threshold level and hazard characteristic information.

[0085] For example, the status parameters (such as pole displacement and actual current value) of various dynamic risks and hazards in the dynamic virtual scene are collected at a frequency of 100ms / time, and the collected parameter values ​​are transmitted to the threshold comparison module. The real-time collected status parameters are compared with the corresponding thresholds at each level in the threshold database to determine the threshold level to which the current parameter belongs: if the parameter is ≥ the warning threshold and < the intervention threshold, it is determined to be a warning level trigger; if the parameter is ≥ the intervention threshold and < the accident threshold, it is determined to be an intervention level trigger; if the parameter is ≥ the accident threshold, it is determined to be an accident level trigger.

[0086] When a certain level of triggering is determined, the system automatically generates a trigger result, the data structure of which includes the threshold level, the type of hazard, the current status parameter value, and the trigger time.

[0087] Step 2.3: Using a branching logic structure, based on the user's operational response efficiency and accuracy in the dynamic virtual scenario, construct emergency scenario subsequences that match the triggering results at each level. Different emergency scenario subsequences contain differentiated hazard evolution paths.

[0088] For example, a branching logic structure is constructed based on two dimensions: the threshold level of the trigger result and the user's operation selection. The user operation selection includes three categories: correct handling operation, partially correct handling operation, and incorrect handling operation. Specific operation definitions refer to industry safety regulations. For each combination of threshold level and operation selection, an emergency scenario subsequence is designed, including the scenario task, feedback logic, and duration limit. Taking a warning-level trigger and correct operation as an example: the emergency scenario subsequence task is to find and mark the risk point. The feedback logic is that the risk state parameters in the scenario stop evolving after correct marking, with a duration limit of 60 seconds. Taking an accident-level trigger and incorrect operation as an example: the emergency scenario subsequence task is to conduct accident rescue (such as simulating the transfer of injured persons), the feedback logic is that the accident's impact range expands in the scenario when an operation is incorrect (such as an increase in the scaffolding collapse area), with a duration limit of 120 seconds.

[0089] By using a scene transition engine, emergency scene subsequences are bound to trigger results and user operation selections, ensuring that after a user is triggered at a certain level, they can enter the matching scene subsequence by performing the corresponding operation.

[0090] Step 2.4: Integrate the emergency scenario subsequences corresponding to the triggering results at each level to form an emergency scenario sequence.

[0091] For example, according to the risk evolution order of early warning level, intervention level, and accident level, the emergency scenario subsequences corresponding to different levels of the same hidden danger type are sorted to form a risk evolution-handling scenario chain for that hidden danger type; for example, the emergency scenario chain for pole displacement risk is an early warning level labeling subsequence - an intervention level adjustment subsequence - an accident level rescue subsequence. All the scenario chains corresponding to the dynamic risks and hidden dangers implanted in step 1.3 are integrated to form an emergency scenario sequence covering multiple hidden danger types and risk levels, stored in a scenario library, and each emergency scenario subsequence is assigned a unique identifier for easy retrieval in subsequent steps.

[0092] Step 3: Based on the trigger result, invoke the corresponding emergency scenario sequence, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through the sensing and interaction device.

[0093] It should be noted that sensor interaction devices refer to a collection of hardware devices used to collect user operation actions and physiological states, including operation interaction devices (operation gloves with integrated electromyography sensors, inertial motion capture devices) and physiological monitoring devices (physiological monitoring wristbands). The devices need to achieve real-time data synchronization with the dynamic virtual scene.

[0094] Operational behavior data refers to quantitative data that describes the characteristics of operational actions generated when users perform operations in an emergency scenario sequence. This includes hand operation trajectory (a three-dimensional coordinate sequence collected by the operating glove), limb movement angle (a joint angle sequence collected by the inertial motion capture device), and equipment operation force (a force value sequence collected by the pressure sensor built into the operating glove).

[0095] Physiological characteristic signals refer to biological signals that reflect the physiological stress state of users in emergency scenario sequences due to scenario stimuli. These include heart rate variability (HRV, which refers to the difference in time interval between consecutive heartbeats), skin conductance response (GSR, which refers to the change in skin surface resistance), and electromyographic activity intensity (EMG, which refers to the intensity of electrical signals generated during muscle contraction). These signals can reflect the user's level of tension and concentration.

[0096] Data correlation with the timeline refers to aligning operational behavior data and physiological characteristic signals with the timeline of the emergency scenario sequence (such as the 1st second or 2nd second after the scenario begins) according to the collection timestamp. This ensures that the operational actions at a certain moment correspond one-to-one with the physiological state at the corresponding moment, providing a data foundation for subsequent correlation analysis.

[0097] This step collects operational behavior data through sensor-interactive devices, which can quantify the standardization of user operations (such as whether hand trajectories conform to standard movements); collect physiological characteristic signals, which can analyze the user's stress response to risks (such as whether the heart rate rises abnormally when facing an accident scenario); and correlate the data with the timeline to solve the problem of disconnect between operation and physiological state, providing complete data support for the subsequent step 4 to analyze the causal relationship between operational abnormalities and physiological fluctuations, and avoiding the one-sidedness of analysis caused by relying solely on operational data.

[0098] In one possible implementation: step 3 includes the following sub-steps:

[0099] Step 3.1: Receive the trigger result generated in Step 2, and call the corresponding level of the scenario in the emergency scenario sequence constructed in Step 2 according to the threshold level in the trigger result.

[0100] For example, the trigger result generated in step 2.2 is received, and the threshold level and hazard type (such as intervention level, electrical overload) are parsed. The scene library is queried based on the threshold level and hazard type to match the corresponding emergency scene sub-sequence identifier. The matched emergency scene sub-sequence is loaded through the scene rendering engine, and the power and data transmission interfaces of the sensing interaction device are simultaneously activated to ensure synchronous startup of the device and the scene (when the scene loading is complete, the device acquisition frequency is synchronously set to 100Hz, consistent with the scene frame rate).

[0101] Step 3.2: Collect user operation behavior data in the corresponding level of emergency scenario through sensor interaction device. The operation behavior data includes hand operation trajectory, limb movement angle and device operation force.

[0102] For example, the user wears an operating glove with an integrated electromyography (EMG) sensor and an inertial motion capture device. After the device is activated, calibration is performed: the operating glove calibrates the zero point of the hand's three-dimensional coordinates by the user performing a fist-clenching and extending motion, while the inertial motion capture device calibrates the zero point of the limb joint angles by the user maintaining a standing, static posture. During the emergency scenario sub-sequence operation, the device collects data at a frequency of 100Hz: it collects the three-dimensional coordinates of each finger joint to form a time series of hand operation trajectories; simultaneously, it collects the force values ​​from the palm pressure sensor on the glove to form a time series of the device's operating force. The inertial motion capture device collects the angle values ​​of the user's upper limb (shoulder, elbow, wrist) and lower limb (hip, knee, ankle) joints to form a time series of limb movement angles.

[0103] Step 3.3 involves collecting the user's physiological characteristic signals in the corresponding level of emergency scenario using a sensor-interactive device. These signals include heart rate variability, skin conductance response, and electromyographic activity intensity. The collection frequency of these physiological characteristic signals is synchronized with the frame rate of the dynamic virtual scene.

[0104] For example, the user wears a physiological monitoring wristband. The wristband collects heart rate signals via photoelectric sensors and skin electrode data on electromyography (EMG) activity. After the device is activated, the user remains still for 5 minutes to calibrate the baseline values ​​of the physiological signals. During the emergency scenario subsequence operation, the physiological monitoring wristband collects data at a frequency of 100Hz. The photoelectric sensor collects the time intervals (RR intervals) of continuous heartbeats, calculates the difference between adjacent RR intervals, and forms a time series of heart rate variability (HRV). The silver-silver chloride electrode on the inside of the wristband collects skin surface resistance values, forming a time series of skin surface resistance (GSR). Electrodes collect electrical signals (range 0-100μV) of the wrist muscles, forming a time series of electromyography (EMG). The timestamps of the physiological characteristic signals and the operational behavior data are unified (based on the scenario activation time, accurate to milliseconds) to ensure consistency in the time dimension of the two types of data.

[0105] HRV (Heart Rate Variability) refers to the difference in time intervals between consecutive heartbeats. It is an important physiological indicator reflecting the functional state of the autonomic nervous system. In this scheme, it is used to quantify the degree of stress response of users when facing risky scenarios. The smaller the HRV fluctuation range, the more stable the user's autonomic nervous regulation and the more moderate the stress state. The larger the fluctuation range, the more likely the user is in a state of high stress such as tension and anxiety.

[0106] GSR (Geodermal Response), also known as electrodermal activity, refers to the fluctuations in skin surface resistance caused by changes in sweat gland secretion. Essentially, it is an external manifestation of sweat gland activity regulated by the autonomic nervous system. In this solution, it is used to help judge the user's emotions and concentration. The smaller the GSR fluctuation range, the more stable the user's emotions and the higher the concentration; the larger the fluctuation range, the more likely the user's emotions will fluctuate due to scene stimuli.

[0107] EMG (Electromyographic Activity Intensity) refers to the intensity of the weak electrical signal generated by muscles during contraction or relaxation. Its value is positively correlated with the activity level of muscle activity. In this protocol, it is used to assess the stability of the user's operation. The smaller the EMG fluctuation range, the more precise the user's muscle control and the more stable the operation. The larger the fluctuation range, the more likely the user's muscle activity is disordered due to tension or lack of skill.

[0108] Step 3.4: Associate the operational behavior data and physiological characteristic signals with the timeline of the corresponding emergency scenario.

[0109] For example, with the start time of the emergency scenario subsequence as time 0, a scenario timeline is established at 10ms intervals, with each time interval corresponding to a time node. Operational behavior data (hand coordinates, joint angles, force values) and physiological characteristic signals (HRV, GSR, EMG) are mapped to the corresponding time nodes according to the acquisition timestamp, forming a correlation table of time nodes, operational data, and physiological data. This correlation table is then bound to the corresponding emergency scenario subsequence identifier and user identity information and stored in a time-series database for subsequent steps 4 and 5.

[0110] Step 4: Synchronize the operational behavior data to the corresponding emergency scenario sequence, guide the user to mark potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operational behavior data and physiological characteristic signals based on the potential risk point marking results to identify dangerous operation modes.

[0111] In one possible implementation: step 4 includes the following sub-steps:

[0112] Step 4.1: Establish a mapping relationship between physical operation parameters and virtual actions in the dynamic virtual scene. Based on this mapping relationship, synchronize the operation behavior data collected in Step 3 to the corresponding level of emergency scene.

[0113] Step 4.2 guides users to mark potential risk points in the corresponding level of emergency scenarios. The marking of potential risk points adopts a combination of three-dimensional coordinate annotation and semantic tags. The semantic tags include the type of hazard, the risk level and the scope of impact.

[0114] Step 4.3: Use time-series comparison method to analyze the correlation between operational behavior data and physiological characteristic signals, and identify the causal relationship between abnormal operational behavior and fluctuations in physiological characteristic signals.

[0115] Step 4.4: Based on the correlation analysis results and the completeness and accuracy of the potential risk point markings, identify hazardous operation modes, including path deviation, response delay, and handling error.

[0116] It should be noted that by establishing a correspondence between physical operation parameters (such as the hand coordinates and force values ​​of the operating gloves) and the motion parameters (such as the coordinates of the virtual hand and the pressure applied by the virtual switch) of virtual operation objects (such as virtual power tools and virtual switches) in the dynamic virtual scene, it is ensured that the user's physical operations can be reflected in the virtual scene in real time. Potential risk point marking refers to the operation of users marking the risk locations in the scene based on the identification of risks in the emergency scenario sequence. The marking method adopts a combination of three-dimensional coordinate annotation and semantic tags, where the three-dimensional coordinate annotation is the spatial location of the risk point (such as the coordinates of the virtual distribution box), and the semantic tags are the attribute information of the risk (hazard type, risk level, and scope of impact).

[0117] Time series alignment methods refer to algorithms used to analyze the similarity and correlation between two time series data (operational behavior data time series and physiological characteristic signal time series). This embodiment uses the dynamic time warping algorithm (DTW), which calculates the distance between the two time series by aligning their time axes. The smaller the distance value, the stronger the correlation. It can be used to identify the associated time periods of operational abnormalities and physiological signal fluctuations.

[0118] Hazardous operation modes refer to the types of operations that users repeatedly perform in emergency scenarios, which do not comply with safety regulations and may exacerbate risks. Based on the characteristics of the operation, they are divided into path deviation type (operation trajectory deviates from the standard path), response delay type (operation execution time exceeds the scenario duration limit), and handling error type (operation content is completely contrary to risk handling requirements).

[0119] This step guides users to mark potential risk points, directly assessing their risk identification capabilities and supplementing insufficient operational data. By comparing and analyzing the correlation between operations and physiology over time, it can uncover the psychological triggers for operational abnormalities (such as operational errors caused by tension). Identifying dangerous operation patterns provides clear problem targets for the subsequent assessment report in step 5 and the cognitive deficit repair in step 6, avoiding problems of untargeted assessments and directionless repairs.

[0120] Step 5: Record the user's three-dimensional motion trajectory in the emergency scenario sequence, and generate a holographic assessment report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes.

[0121] In one possible implementation: step 5 includes the following sub-steps:

[0122] Step 5.1: Record the user's three-dimensional motion trajectory in the corresponding level of emergency scenario using a spatial positioning device.

[0123] It should be noted that spatial positioning equipment refers to devices used to collect the overall movement trajectory of a user in an emergency scenario sequence. These devices employ multi-base station optical positioning systems (such as a positioning network consisting of four positioning base stations) to capture the three-dimensional coordinates of positioning markers worn by the user (such as reflective markers attached to the head and torso) to generate the user's overall movement trajectory. The positioning accuracy is ≤1mm. The three-dimensional movement trajectory refers to the three-dimensional coordinate sequence of the user's overall body position changing over time in the emergency scenario sequence, including the head trajectory (reflecting the user's observation path) and the torso trajectory (reflecting the user's movement path). This can be used to analyze whether the user has missed key scene areas (such as failing to observe the area near the electrical distribution box).

[0124] Step 5.2: Extract the operational behavior data and physiological characteristic signals collected in Step 3, the potential risk point marking results and dangerous operation modes generated in Step 4, and construct a multi-dimensional analysis dataset.

[0125] It should be noted that the multidimensional analysis dataset refers to a collection that integrates five types of data: three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes. The data must be aligned according to the time axis, and each data item must include attributes such as data type, collection time, data value, and whether it is abnormal. This is the basis for quantitative processing.

[0126] Step 5.3: Quantify the multi-dimensional analysis dataset to generate quantitative results for operational compliance score, risk identification accuracy, and physiological stress index.

[0127] It should be noted that the quantitative processing results refer to the indicators obtained after quantitatively scoring various types of data in the multi-dimensional analysis dataset, including operational compliance score (based on the standardization of operational behavior data), risk identification accuracy (based on the results of potential risk point marking), and physiological stress index (based on the degree of fluctuation of physiological characteristic signals). Each indicator is scored out of 100, with higher scores indicating better performance.

[0128] For example, the scoring criteria for operational compliance (out of 100 points) are as follows: the weight of operational path deviation is 40%, with a deviation value ≤10mm earning 40 points, 10mm < deviation value ≤20mm earning 20 points, and a deviation value >20mm earning 0 points; the weight of operational force compliance is 30%, with force within the standard range earning 30 points, otherwise earning 0 points; the weight of timely operational response is 30%, with a response delay ≤10 seconds earning 30 points, 10 seconds < delay ≤20 seconds earning 15 points, and a delay >20 seconds earning 0 points.

[0129] Operational compliance score = (path deviation score × 40%) + (force compliance score × 30%) + (timeliness of response score × 30%).

[0130] The scoring criteria for risk identification accuracy (out of 100 points) are as follows: the weight of the marker position accuracy is 50%, with a position deviation ≤ 10mm earning 50 points, 10mm < deviation ≤ 20mm earning 25 points, and deviation > 20mm earning 0 points; the weight of the label matching degree is 50%, with a perfect label match earning 50 points, a partial match earning 25 points, and a complete mismatch earning 0 points.

[0131] Risk identification accuracy score = (location accuracy score × 50%) + (label matching score × 50%).

[0132] Physiological Stress Index (maximum score 100, higher score indicates more stable stress state):

[0133] The scoring criteria are as follows: HRV stability has a weight of 40%, with HRV fluctuation range ≤ 20ms scoring 40 points, fluctuation range < 40ms scoring 20 points, and fluctuation range > 40ms scoring 0 points; GSR stability has a weight of 30%, with GSR fluctuation range ≤ 10kΩ scoring 30 points, fluctuation range < 20kΩ scoring 15 points, and fluctuation range > 20kΩ scoring 0 points; EMG stability has a weight of 30%, with EMG fluctuation range ≤ 10μV scoring 30 points, fluctuation range < 20μV scoring 15 points, and fluctuation range > 20μV scoring 0 points.

[0134] Physiological stress index = (HRV stability score × 40%) + (GSR stability score × 30%) + (EMG stability score × 30%).

[0135] Step 5.4: Integrate the three-dimensional motion trajectory, quantitative processing results, and dangerous operation modes to generate a holographic assessment report containing visual analysis content.

[0136] It should be noted that the holographic assessment report refers to a comprehensive assessment document that includes "data visualization, quantitative indicators, and problem analysis." Data visualization includes three-dimensional motion trajectory playback and operation-physiological correlation curves; quantitative indicators include various percentage scores; and problem analysis includes a detailed description of dangerous operation modes and inferences about their causes, which serves as the direct basis for the subsequent construction of a cognitive deficit map.

[0137] This step supplements the analysis of the user's overall movement by recording the three-dimensional motion trajectory, avoiding evaluation omissions caused by focusing only on local operations; it constructs a multi-dimensional dataset to ensure that the evaluation is based on complete data and avoids one-sidedness; quantitative processing transforms qualitative data (such as labeled results) into quantitative indicators (such as accuracy scores) to solve the problem of subjective evaluation; and the holographic evaluation report integrates scattered data and analysis into a structured document, providing a clear and directly usable basis for the construction of the cognitive deficit map in step 6, ensuring the targeted nature of cognitive repair.

[0138] Step 6: Construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, use the scenario backtracking tool to analyze the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in Step 2. Conduct targeted re-experience training in this emergency scenario sequence.

[0139] It should be noted that the user cognitive deficiency map refers to a structured map constructed based on the problem analysis in the holographic assessment report, describing the user's cognitive shortcomings in safety training. It includes cognitive dimensions, cognitive shortcomings, and shortcoming weights. The scenario backtracking tool refers to a software tool used to freeze and analyze key frames of emergency scenario sequences. It allows users to manually select the scenario runtime segment (e.g., seconds 10-15), freeze the scene frames (static images) for that period, and query the safety parameters of virtual components in that frame. The tool needs to be connected to the emergency scenario sequence database in real time. The safety parameters of virtual components refer to the technical parameters of virtual building components or equipment in the emergency scenario sequence used to determine their safety status, including structural load-bearing limits, equipment operating thresholds, and the effective range of protective facilities. These parameters are the core basis for determining whether the risk exceeds the standard. Targeted re-experience training refers to reproducing related scenarios in the corresponding emergency scenario sequence for cognitive shortcomings with high weights in the user cognitive deficiency map, designing targeted training tasks, and repeatedly training to repair cognitive deficiencies. After training, the repair effect needs to be verified.

[0140] This step transforms operational issues in the assessment report into cognitive problems by constructing a user cognitive deficiency map, addressing the fundamental problem of existing training that only addresses operational aspects without addressing cognitive ones. A scenario retrospective tool analyzes safety parameters, enabling trainees to understand the root causes of excessive risks (such as pole displacement exceeding the bearing limit leading to collapse), rather than simply recognizing operational errors. Targeted re-experience training designs tasks to address cognitive shortcomings, achieving precise repair and avoiding untargeted repetitive training. Simultaneously, training results are fed back to step 1 to optimize the complexity of subsequent scenarios, forming a training closed loop.

[0141] In one possible implementation: step 6 includes the following sub-steps:

[0142] Step 6.1: Based on the quantitative processing results and dangerous operation modes in the holographic assessment report, construct a user cognitive deficiency map. The user cognitive deficiency map includes cognitive shortcomings in risk identification, handling logic, and emergency response dimensions. The degree of cognitive shortcomings in each dimension is quantified by weighting.

[0143] For example, based on the core competency requirements of construction safety training, three cognitive dimensions are defined: risk identification, handling logic, and emergency response. The cognitive weaknesses in each dimension are extracted from the problem analysis of the holographic assessment report: For the risk identification dimension, if the risk identification accuracy score is ≤50 points, or if key areas are missed (e.g., insufficient identification of scaffolding upright displacement risk, incorrect judgment of electrical overload risk level). For the handling logic dimension, if there are erroneous handling operation patterns, or if the operation compliance score is ≤50 points, weaknesses are extracted (e.g., disordered sequence of scaffolding upright adjustments, failure to disconnect power before handling electrical overload). For the emergency response dimension, if there are delayed response operation patterns, or if the physiological stress index is ≤50 points, weaknesses are extracted (e.g., psychological tension in accident scenarios leading to delayed emergency operations, unclear memory of emergency rescue procedures).

[0144] The Analytic Hierarchy Process (AHP) was used to assign weights to each cognitive weakness. A judgment matrix was constructed, and 3-5 construction safety experts were invited to compare the severity of each weakness pairwise. The weight of each weakness was calculated using the judgment matrix (the sum of the weights is 1). A weakness with a weight ≥ 0.3 was considered a high-weight weakness (requiring priority for repair), a weakness with a weight ≤ 0.1 < 0.3 was considered a medium-weight weakness, and a weakness with a weight < 0.1 was considered a low-weight weakness.

[0145] The cognitive dimensions, cognitive shortcomings, and the weight of the shortcomings are integrated into a visual user cognitive deficiency map (using mind map format). The map uses different colors to mark the weight level (red indicates high weight, yellow indicates medium weight, and blue indicates low weight) and is stored in the database.

[0146] Step 6.2: Based on the cognitive shortcomings with higher weights in the user cognitive deficiency map, locate the multi-level risk trigger threshold level and emergency scenario sequence corresponding to Step 2.

[0147] For example, for high-weight shortcomings in the cognitive deficit map, we can analyze their corresponding risk types and risk levels: For instance, a high-weight shortcoming is insufficient identification of pole displacement risk, the corresponding risk type is pole displacement risk, and the risk level is intervention level (because the shortcoming is most obvious in intervention level scenarios).

[0148] Based on the risk type and risk level, query the emergency scenario sequence library from step 2 to locate the corresponding emergency scenario sub-sequence identifier, and simultaneously record the multi-level risk trigger thresholds corresponding to that scenario. Load the located emergency scenario sub-sequence into the scenario cache area, and retrieve the corresponding multi-level risk trigger threshold parameters from the threshold database of step 2, preparing for subsequent parsing of safety parameters and re-experience training.

[0149] Step 6.3: Freeze the keyframes of the target emergency scenario sequence using the scenario backtracking tool, and analyze the safety parameters of the virtual components in the target emergency scenario sequence. The safety parameters of the virtual components include the structural bearing limit, equipment operating threshold, and effective range of protective facilities.

[0150] For example, launch the scene backtracking tool, load the emergency scene sub-sequence located in step 6.2, and play it to the period when the shortcoming occurred (e.g., seconds 10-15, the period when the user did not identify the risk of pole displacement). Manually click the freeze button to freeze the scene frame for that period (e.g., the scene frame at second 12, at which time the pole displacement is 18mm). In the frozen scene frame, click on the target virtual component (e.g., the pole). The scene backtracking tool automatically sends a parameter query request to the real-time database of the emergency scene sequence. The database returns the safety parameters of the component: the maximum axial bearing capacity of the pole (e.g., 30kN) and the maximum allowable displacement (e.g., 15mm, i.e., the displacement corresponding to the intervention level threshold). The tool automatically calculates the difference between the actual displacement (18mm) and actual force (25kN) of the pole in the current scene frame and the safety parameters (displacement exceeds the tolerance by 3mm, force does not exceed the tolerance).

[0151] The tool pops up a parameter display panel on the right side of the scene frame, displaying parameters in the format of safety parameters (red), actual parameters (blue), and differences (green). It also provides industry standard references for the parameters to help users understand the source and meaning of the parameters.

[0152] Step 6.4: Reproduce the associated hidden danger scenarios in the emergency scenario sequence of the corresponding level, conduct targeted re-experience training, and verify the effect of correcting cognitive shortcomings through the training process.

[0153] For example, a targeted training task is designed for high-weight shortcomings: If the shortcoming is insufficient risk identification: the task is to find and mark 3 pole displacement risk points in the scene, check the safety parameters after marking, understand the reasons for exceeding the tolerance, the task duration is 90 seconds, and parameter prompts pop up every 10 seconds in the scene.

[0154] If the weakness is a confused handling logic: the task is to adjust the out-of-tolerance poles according to the standard procedure, and after each step, it is necessary to confirm whether the safety parameters meet the standards. Error process prompts should be set in the task.

[0155] The user performs training tasks in the emergency scenario subsequence located in step 6.2, and the scenario backtracking tool records the user's operation data and parameter viewing records in real time (such as the number of times parameters are viewed and the duration of each viewing).

[0156] After training, the quantitative indicators from step 5 are used for verification: Risk identification dimension: Risk identification accuracy needs to be improved to ≥70 points, and key area coverage needs to be improved to ≥90%. Handling logic dimension: Operational compliance score needs to be improved to ≥70 points, and the incidence of dangerous operation modes needs to be reduced to ≤10%. Emergency response dimension: Physiological stress index needs to be improved to ≥70 points, and operational response delay needs to be reduced to ≤10 seconds. If the training fails, repeat the training task in step 6.4 until verification is successful.

[0157] Example 2 provides a VR-based construction experience system, which applies the above-mentioned VR-based construction experience method, including:

[0158] A dynamic virtual scene construction module is used to generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model through machine learning algorithms and embed dynamic risks and hidden dangers to form a dynamic virtual scene.

[0159] The risk threshold setting and emergency scenario construction module is used to set multi-level risk trigger thresholds based on dynamic virtual scenarios. The multi-level risk trigger thresholds correspond to different danger levels in the dynamic virtual scenarios. When the state parameters of dynamic risk hazards reach the corresponding level of multi-level risk trigger thresholds, a trigger result is generated. An emergency scenario sequence matching the trigger results at each level is constructed using a branch logic structure.

[0160] A multimodal data acquisition module is used to call the corresponding emergency scenario sequence according to the trigger result, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through a sensing and interaction device;

[0161] The operation synchronization and hazard pattern recognition module is used to synchronize operation behavior data to the corresponding emergency scenario sequence, guide the user to complete the marking of potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operation behavior data and physiological characteristic signals by combining the potential risk point marking results to identify dangerous operation patterns.

[0162] The three-dimensional trajectory recording and holographic evaluation module is used to record the user's three-dimensional motion trajectory in an emergency scenario sequence, and generate a holographic evaluation report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes.

[0163] The cognitive deficiency analysis and targeted training module is used to construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, the module analyzes the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in the risk threshold setting and emergency scenario construction module through the scenario backtracking tool, and conducts targeted re-experience training in the emergency scenario sequence.

[0164] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A construction experience method based on VR technology, characterized in that, Includes the following steps: Step 1: Generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model and embed dynamic risks and hidden dangers through machine learning algorithms to form a dynamic virtual scene; Step 1 includes the following sub-steps: Step 1.1: Collect 3D point cloud data, building information model and historical accident case data of each stage of building construction, and fuse the above data based on digital twin technology to generate a basic scene model of building construction. Step 1.2, determine whether there are user historical operation records: if there are, use machine learning algorithms to parse the user historical operation records, establish a matching model between scene complexity and user operation ability, and adjust the scene complexity of the building construction basic scene model based on the matching model; if there are no such records, use preset scene complexity parameters to adjust the scene complexity of the building construction basic scene model. Step 1.3: Incorporate dynamic risks and hazards into the adjusted building construction basic scenario model. Dynamic risks and hazards are potential safety risks that change state with time parameters or user operation intervention. Step 1.4: Integrate and adjust the basic construction scenario model and dynamic risks and hazards after adjusting the scenario complexity to form a dynamic virtual scenario; Step 2: Set multi-level risk trigger thresholds based on dynamic virtual scenarios. The multi-level risk trigger thresholds correspond to different danger levels in the scenario. When the state parameters of dynamic risk hazards reach the corresponding level of multi-level risk trigger thresholds, a trigger result is generated. An emergency scenario sequence matching the trigger results at each level is constructed using a branch logic structure. Step 2 includes the following sub-steps: Step 2.1: Based on the scene parameters of the dynamic virtual scene, set multi-level risk trigger thresholds. The multi-level risk trigger thresholds include the early warning threshold for the initial state of the corresponding hidden danger, the intervention threshold for the evolution state of the corresponding hidden danger, and the accident threshold for the failure state of the corresponding hidden danger. Step 2.2: Monitor the status parameters of dynamic risks and hazards in the dynamic virtual scene in real time. When the status parameters of dynamic risks and hazards reach any level of the multi-level risk trigger threshold, generate a trigger result containing the threshold level and hazard characteristic information. Step 2.3: Using a branch logic structure, based on the user's operational response efficiency and operational accuracy in the dynamic virtual scene, construct emergency scenario subsequences that match the trigger results at each level. Different emergency scenario subsequences contain differentiated hazard evolution paths. Step 2.4: Integrate the emergency scenario subsequences corresponding to the triggering results at each level to form an emergency scenario sequence; Step 3: Based on the trigger result, invoke the corresponding emergency scenario sequence, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through the sensing and interaction device; Step 4: Synchronize the operational behavior data to the corresponding emergency scenario sequence, guide the user to mark potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operational behavior data and physiological characteristic signals based on the potential risk point marking results to identify dangerous operation modes. Step 5: Record the user's three-dimensional motion trajectory in the emergency scenario sequence, and generate a holographic assessment report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes; Step 6: Construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, use the scenario backtracking tool to analyze the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in Step 2. Conduct targeted re-experience training in this emergency scenario sequence. Step 6 includes the following sub-steps: Step 6.1: Based on the quantitative processing results and dangerous operation modes in the holographic assessment report, construct a user cognitive deficiency map. The user cognitive deficiency map includes cognitive shortcomings in risk identification, handling logic, and emergency response dimensions. The degree of cognitive shortcomings in each dimension is quantified by weighting. Step 6.2: Based on the cognitive shortcomings with higher weights in the user cognitive deficiency map, locate the multi-level risk trigger threshold level and emergency scenario sequence corresponding to Step 2; Step 6.3: Freeze the keyframes of the target emergency scenario sequence using the scenario backtracking tool, and analyze the safety parameters of the virtual components in the target emergency scenario sequence. The safety parameters of the virtual components include the structural bearing limit, equipment operating threshold, and effective range of protective facilities. Step 6.4: Reproduce the associated hidden danger scenarios in the emergency scenario sequence of the corresponding level, conduct targeted re-experience training, and verify the effect of correcting cognitive shortcomings through the training process.

2. The construction experience method based on VR technology according to claim 1, characterized in that: Step 3 includes the following sub-steps: Step 3.1: Receive the trigger result generated in Step 2, and call the corresponding level of the scenario in the emergency scenario sequence constructed in Step 2 according to the threshold level in the trigger result; Step 3.2: Collect user operation behavior data in the corresponding level of emergency scenario through sensor interaction device. The operation behavior data includes hand operation trajectory, limb movement angle and device operation force. Step 3.3: Collect physiological characteristic signals of users in the corresponding level of emergency scenario through sensor interaction device. The physiological characteristic signals include heart rate variability, skin conductance response and electromyographic activity intensity. The collection frequency of physiological characteristic signals is synchronized with the frame rate of dynamic virtual scene. Step 3.4: Associate the operational behavior data and physiological characteristic signals with the timeline of the corresponding emergency scenario.

3. The construction experience method based on VR technology according to claim 1, characterized in that: Step 4 includes the following sub-steps: Step 4.1: Establish a mapping relationship between physical operation parameters and virtual actions in the dynamic virtual scene, and based on this mapping relationship, synchronize the operation behavior data collected in Step 3 to the corresponding level of emergency scene; Step 4.2 guides users to mark potential risk points in the corresponding level of emergency scenarios. The marking of potential risk points adopts a combination of three-dimensional coordinate annotation and semantic tags. The semantic tags include the type of hazard, risk level and scope of impact. Step 4.3: Use time-series comparison method to analyze the correlation between operational behavior data and physiological characteristic signals, and identify the causal relationship between abnormal operational behavior and fluctuations in physiological characteristic signals; Step 4.4: Based on the correlation analysis results and the completeness and accuracy of the potential risk point markings, identify hazardous operation modes, including path deviation, response delay, and handling error.

4. The construction experience method based on VR technology according to claim 1, characterized in that: Step 5 includes the following sub-steps: Step 5.1: Record the user's three-dimensional movement trajectory in the corresponding level of emergency scenario using spatial positioning equipment; Step 5.2: Extract the operational behavior data and physiological characteristic signals collected in Step 3, the potential risk point marking results and dangerous operation patterns generated in Step 4, and construct a multi-dimensional analysis dataset; Step 5.3: Quantify the multi-dimensional analysis dataset to generate quantitative results for operational compliance score, risk identification accuracy and physiological stress index; Step 5.4: Integrate the three-dimensional motion trajectory, quantitative processing results, and dangerous operation modes to generate a holographic assessment report containing visual analysis content.

5. A construction experience system based on VR technology, characterized in that, A construction experience method based on VR technology according to any one of claims 1-4 includes: A dynamic virtual scene construction module is used to generate a basic construction scene model based on digital twin technology, adjust the scene complexity of the basic construction scene model through machine learning algorithms and embed dynamic risks and hidden dangers to form a dynamic virtual scene. The risk threshold setting and emergency scenario construction module is used to set multi-level risk trigger thresholds based on dynamic virtual scenarios. The multi-level risk trigger thresholds correspond to different danger levels in the dynamic virtual scenarios. When the state parameters of dynamic risk hazards reach the corresponding level of multi-level risk trigger thresholds, a trigger result is generated. An emergency scenario sequence matching the trigger results at each level is constructed using a branch logic structure. A multimodal data acquisition module is used to call the corresponding emergency scenario sequence according to the trigger result, and collect the user's operation behavior data and physiological characteristic signals in the emergency scenario sequence through a sensing and interaction device; The operation synchronization and hazard pattern recognition module is used to synchronize operation behavior data to the corresponding emergency scenario sequence, guide the user to complete the marking of potential risk points in the emergency scenario sequence that match the corresponding level of multi-level risk trigger thresholds, and analyze the correlation between operation behavior data and physiological characteristic signals by combining the potential risk point marking results to identify dangerous operation patterns. The three-dimensional trajectory recording and holographic evaluation module is used to record the user's three-dimensional motion trajectory in an emergency scenario sequence, and generate a holographic evaluation report based on the three-dimensional motion trajectory, operational behavior data, physiological characteristic signals, potential risk point marking results, and dangerous operation modes. The cognitive deficiency analysis and targeted training module is used to construct a user cognitive deficiency map based on the holographic assessment report. Based on the user cognitive deficiency map, the module analyzes the safety parameters of virtual components in the emergency scenario sequence corresponding to the multi-level risk trigger thresholds associated with user cognitive deficiencies in the risk threshold setting and emergency scenario construction module through the scenario backtracking tool, and conducts targeted re-experience training in the emergency scenario sequence.

Images