Operation specification evaluation method and system based on VR safety training system

Abstract

The invention discloses an operation specification evaluation method and system based on a VR safety training system, relates to the technical field of electric power safety, and solves the problem that the operation specification degree of a student cannot be subjected to atomic-scale, multi-dimensional and procedural fine evaluation in the prior art. According to the method, firstly, a general safety regulation is deconstructed into a'action-object-constraint condition 'triple, and a knowledge graph is constructed, so that a system can identify key details such as'a locking anti-falling device must hear clicking sound and slightly pull verification' and the like instead of only seeing whether the action is completed or not; secondly, respectively quantizing space trajectory precision, visual inspection sufficiency and action sequence preciseness by applying algorithms such as dynamic time warping, an illustrative function and an editing distance; finally, differentiated processing is carried out on violation and slight deviation through a comprehensive scoring mechanism containing a rigid penalty term, a traceable, comparable and forcibly improved quantitative evaluation system is formed, and safety capability evaluation really has refined, standardized and executable industry values.

Description

Technical Field

[0001] This invention relates to the field of power safety technology, and in particular to an operational procedure assessment method and system based on a VR safety training system. Background Technology

[0002] Currently, with the rapid growth of offshore wind power installed capacity, the multiple risks faced by its operation and maintenance operations, such as falls from heights, electric shocks, fires, and severe sea conditions, pose unprecedented challenges to safety training. Traditional training methods relying on centralized lectures, paper textbooks, and on-site explanations are no longer sufficient to meet the demands of modern offshore wind farms for high coverage, high effectiveness, and traceability of contractor personnel. In recent years, the maturity of cloud computing, virtual reality (VR), and digital twin technologies has provided technical support for building highly simulated offshore operating environments, making it possible to simulate high-risk scenarios such as typhoons, surges, and equipment failures in virtual space. In particular, given the high turnover rate and weak safety foundation of contractor personnel, the industry urgently needs to shift safety training from "knowledge transmission" to "precise assessment of operational behavior," using quantitative data to achieve accurate profiling and access control of personnel capabilities. This has become the core technological direction for the digital transformation of offshore wind power safety production.

[0003] However, existing VR safety training systems suffer from fundamental shortcomings: they cannot provide atomic-level, multi-dimensional, and process-oriented detailed assessments of trainees' operational compliance. Current technologies largely remain at a coarse-grained "result-oriented" level, such as "whether the designated location has been reached" or "whether the fire has been extinguished," failing to identify crucial details such as "whether a click was heard and gently pulled to verify when the fall arrestor is locked," "whether the hydraulic oil level is checked for more than 3 seconds," and "whether the power-off operation strictly follows the sequence of power off-voltage-voltage verification-grounding wire connection." This blind spot in assessment is particularly serious in offshore wind power scenarios—unverified locking could lead to falls from hundreds of meters high, insufficiently checked equipment status could cause fires, and incorrectly sequenced electrical operations could directly cause electrocution. Because these fatal details cannot be captured, training easily becomes a "pass-the-test" exercise, especially for newly arrived contractors with weak safety foundations, making it difficult to truly improve their risk identification and emergency response capabilities.

[0004] Therefore, there is a need for an operational specification assessment method and system based on VR safety training systems. Summary of the Invention

[0005] To address the problem that existing technologies cannot provide precise, multi-dimensional, and process-oriented assessments of trainees' operational compliance, this invention provides an operational compliance assessment method and system based on a VR safety training system. This method deconstructs operational procedures into constrained atomic action units, integrates multimodal data such as controller posture DTW algorithm, eye-tracking gaze rate analysis, and edit distance temporal evaluation, and constructs a comprehensive scoring system combining dynamic weights and rigid penalties. This achieves precise quantification and real-time feedback for every operational detail. The specific technical solution is as follows: An operational procedure evaluation method based on a VR safety training system includes the following steps: The safety operating procedures are atomically decomposed into action-object-constraint triples to construct a knowledge graph. Data on the handgrip posture trajectory, eye-tracking fixation points, and action timing of trainees' operations are collected. Posture deviation is calculated using a dynamic time warping algorithm, fixation rate in key areas is calculated using an indicator function, and timing compliance is calculated using edit distance. All of these are then integrated into a compliance index. Calculate the comprehensive normative score, which includes rigid penalty items.

[0006] Preferably, constructing a knowledge graph specifically includes the following steps: Break down safety procedures into sets of atomic actions. ; Set of atomic actions Import the graph database and define the action relationships and weights between them: in, To standardize knowledge graph objects, they must contain all the information of a directed graph; , represents the set of vertices of a graph, where each vertex... Represents an atomic action; Represents the set of edges of a graph; For two atomic actions that have a dependency relationship, This is a preliminary action. For subsequent actions; The weights are assigned based on the risk consequences classification.

[0007] Preferably, the attitude deviation is calculated in the following way: in, The three-dimensional trajectory sequence of the handle operated by the student. ; This is the trajectory sequence of the technician's standard operating procedure. ; , indicating time The three-dimensional coordinates of the handle; This provides a set of all possible timing alignment paths to address the issue of inconsistent student operation speeds with the standard. For the alignment function, the first element of the standard sequence is... Frame mapped to the first of the student sequence frame, Given the Euclidean norm, calculate the straight-line distance between two points in space. The length of the trajectory sequence; The posture score is obtained: in, This represents the spatial deviation tolerance threshold; exceeding this value is considered a serious deviation. This represents the weighting coefficient of the pose dimension in the total score.

[0008] Preferably, the fixation rate in key areas is calculated in the following way: in, This indicates the compliance rate. This indicates the total operation time of the action; This is an indicator function that returns either 0 or 1. For a moment The three-dimensional coordinates of the eye's gaze point; A 3D bounding box representing a key region; The minimum fixation duration required by the standard; Visual compliance points were deducted. in, This represents the weighting coefficient of the visual dimension in the total score.

[0009] Preferably, timing compliance is calculated in the following way: in, This represents the sequence of actions actually performed by the trainee; This represents the sequence of actions required by the specification. Represents the set of all possible sequences of edit operations; The k-th editing operation Indicates the cost of editing operations; Indicates the total length of the edit sequence; Resulting in time series scores: in, Indicates the length of the standard action sequence (used for normalization); This represents the weighting coefficient of the time series dimension in the total score.

[0010] Preferably, the compliance index is calculated using the following formula: in, For the first The specification conformity index of an atomic action Score the pose dimension; Scoring is based on the time series dimension; Score the visual dimension; Tool usage score; Indicates the weight.

[0011] The preferred method for calculating the comprehensive standard score is as follows: in, This represents the total number of actions to be performed in the standard action sequence. This is the penalty coefficient; For the first The immediate standardization index of each action; This represents the dynamic weight of the action in scene R; This is an indicator function; it is 1 if the action is actually performed, and 0 otherwise. The total weight of all actions that should be performed; This refers to the number of serious violations.

[0012] An operational procedure assessment system based on a VR safety training system, applied to the method described above, includes: Data acquisition unit: Connects to the VR safety training system to acquire various data during the trainee's operation in real time, and then transmits the data to the data processing and scoring output unit for processing; Data processing and scoring output unit: Using the method of Example 1 or Example 2, the long-term compliance index and dynamic baseline are finally output, and the long-term compliance index and dynamic baseline are transmitted to the decision-making unit; Decision-making unit: The decision-making unit makes decisions on colleges or contractors based on long-term standard implementation index and dynamic baseline.

[0013] A computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the operation specification assessment method of the VR-based safety training system as described above.

[0014] A processor for running a program, wherein the program, when running, executes the operation specification assessment method for a VR-based safety training system as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention addresses the critical deficiency of existing VR safety training, which fails to provide atomic-level, multi-dimensional, and process-oriented detailed assessments of operational procedures. This technical solution achieves a precise breakthrough through three innovations: First, it deconstructs general safety procedures into "action-object-constraint" triplets and constructs a knowledge graph, enabling the system to recognize key details such as "the locking of the fall arrestor must be heard with a click and gently pulled for verification," rather than simply focusing on whether the action is completed. Second, it simultaneously collects multimodal data such as handle posture, eye movement, and operation sequence, using algorithms such as dynamic time warping, indicative functions, and edit distance to quantify spatial trajectory accuracy, visual inspection sufficiency, and the rigor of action sequence, then integrates these into a compliance index, achieving a paradigm shift from "results-oriented" to "process-diagnostic." Finally, through a comprehensive scoring mechanism including rigid penalties, it differentiates between fatal violations and minor deviations, forming a traceable, comparable, and enforceable quantitative assessment system. This completely solves the core pain points of traditional training—namely, the lack of control over contractor personnel's operational details and insufficient identification of risk chain reactions—making safety capability assessment truly refined, standardized, and executable within the industry. Attached Figure Description

[0016] (none). Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0021] In one embodiment of the present invention, an operational specification evaluation method based on a VR safety training system is provided, comprising the following steps: Step 1: Decompose the procedure into atomic action units and establish a standard knowledge graph and atomic action library. Details are as follows: S101: By organizing a safety expert group, site manager, senior operation and maintenance workers, etc., we disassembled the "Safety Regulations for Offshore Wind Farms" DL / T 796 line by line to identify all the indivisible actions in the core processes such as wind turbine tower climbing, submarine cable splicing, and fire emergency response. S102: Establish standardized data descriptions for each atomic action, including: action verbs (locking / checking / wearing), operation objects (fall arrestor / hydraulic valve / life vest), and constraints (auditory confirmation: whether a "click" is heard / visual inspection for 3 seconds / torque meets the standard), forming a structured library of atomic actions. .

[0022] S103: Import the action library into the graph database, and define the action relationships and weights between actions: in, To standardize knowledge graph objects, they must contain all the information of a directed graph; , represents the set of vertices of a graph, where each vertex... Represents an atomic action (such as "locking the fall arrestor" or "visually checking the oil level"). Represents the set of edges of a graph; For two atomic actions that are dependent on each other ( This is a preliminary action. (for subsequent actions) The weights are assigned based on the risk consequences classification.

[0023] For example, This indicates that the power must be turned off before touching it, otherwise the fatal risk weight is 10.

[0024] For example, the more serious the risk consequences, the higher the score. If the violation leads to fatal risks such as fire / electric shock, the weight is set to 10. If the violation leads to major risks such as falls / collisions, the weight is set to 5. If the violation leads to minor injuries or equipment damage, the weight is set to 5.

[0025] In other words, by configuring different weights, subjective risk perceptions are transformed into objective numerical weights. In offshore wind power accidents, the difference in harm between "touching high-voltage terminals without disconnecting power" and "not wearing gloves" is addressed through weighting. By quantifying and differentiating, algorithmic resources are prioritized for monitoring high-risk action chains.

[0026] Step 2: Multimodal Behavior Capture and Atomic Action-Level Evaluation S201: Perform a compliance check on the handle posture, as follows: First, the OptiTrack optical tracking system is deployed in the VR training cabin with a sampling rate of 120Hz to ensure that the three-dimensional position error of the tool held by the trainee is less than 2 mm and the rotation accuracy is less than 1 degree.

[0027] Then, standard operations (such as locking the fall arrestor and using a torque wrench) by three technicians with more than 10 years of experience were recorded to generate a standard attitude trajectory library. P 标准 Each action samples 50-80 key points.

[0028] Finally, during the student's operation, the system uses a 100-millisecond sliding window to calculate the dynamic time warping (DTW) distance between the current trajectory and the standard trajectory in real time. in, The three-dimensional trajectory sequence of the handle operated by the student. ; This is the trajectory sequence of the technician's standard operating procedure. ; , indicating time The three-dimensional coordinates of the handle; This provides a set of all possible timing alignment paths to address the issue of inconsistent student operation speeds with the standard. For the alignment function, the first element of the standard sequence is... Frame mapped to the first of the student sequence frame, Given the Euclidean norm, calculate the straight-line distance between two points in space. is the length of the trajectory sequence.

[0029] DTW allows trainees to operate at varying speeds, but strictly assesses the accuracy of the spatial trajectory. When using a torque wrench to tighten bolts in offshore wind power, the standard trajectory is a uniform circular motion. DTW can identify whether trainees are engaging in dangerous operations such as "swaying and striking".

[0030] Based on this, the posture score is derived: in, This represents the spatial deviation tolerance threshold; exceeding this value is considered a serious deviation. This represents the weighting coefficient of the pose dimension in the total score.

[0031] S202: Perform standardized verification of eye-tracking.

[0032] First, select a VR headset with an integrated eye tracker (such as HTC Vive Pro Eye) with a sampling rate of no less than 120Hz to calibrate the eye movement model for each student.

[0033] Then, mark key inspection areas (ROIs) on the 3D equipment model, such as hydraulic level gauges, emergency stop buttons, pressure relief valves, etc., and set a standard look-through time T standard (e.g., 3 seconds) for each ROI.

[0034] While the trainee performs the examination, the system calculates in real time the percentage of time the eye spends within the ROI: in, This represents the compliance rate, a dimensionless value between 0 and 1, with the closer to 1 being better. This indicates the total operation time of the action; This is an indicator function that returns either 0 or 1. It returns 1 when the condition is met (the eye is within the ROI), otherwise 0. For a moment The three-dimensional coordinates of the eye's gaze point; A three-dimensional bounding box representing a critical area (e.g., a 0.1m × 0.1m × 0.05m area around a hydraulic level gauge). The minimum gaze duration required by the standard (e.g., 3 seconds) is predefined by the safety procedure.

[0035] Based on this, the visual compliance points are deducted as follows: in, This represents the weighting coefficient of the visual dimension in the total score.

[0036] During offshore wind turbine inspections, failing to monitor the hydraulic oil level indicator and starting the turbine directly may lead to major malfunctions. Ensure that visual inspection actions are performed correctly.

[0037] S203: Evaluate the operation sequence.

[0038] First, based on the action dependencies in the knowledge graph, a standard action sequence for each training scenario is generated. For example, a fire emergency drill includes 23 sequential actions. Then, as each action is completed by the trainee in the VR environment, the system records its timestamp. t ( a i ), construct the actual execution sequence The system detects omissions, duplicates, or out-of-order sequences in real time. It uses Levenshtein to edit distance quantization timing deviations. in, This represents the sequence of actions actually performed by the trainee; This represents the sequence of actions required by the specification. Represents the set of all possible sequences of editing operations (insert, delete, swap); The kth editing operation (e.g., "delete") a 断电 ") The cost of editing operations is determined by security. Indicates the total length of the edit sequence; For example, we define the insertion / deletion cost as c(·) = 10 and the exchange cost as c(·) = 5. That is, LD ≥ 20 is equivalent to missing 2 critical actions.

[0039] Based on this, the time series score is derived: in, Indicates the length of the standard action sequence (used for normalization); This represents the weighting coefficient of the time series dimension in the total score.

[0040] In an emergency response to a fire at an offshore wind power plant, if trainees skip "cutting off the power to the wind turbine" and go directly to extinguish the fire, the LD value will increase sharply, causing the timing score to approach zero.

[0041] Step 3: Dynamic weight fusion and comprehensive score generation S301: Synthesis of Four-Dimensional Real-Time Normativity Index.

[0042] For each atomic action completed by the trainee (such as locking the fall arrestor), the system calculates their posture score, timing score, visual score, and tool usage score in parallel within 0.1 seconds. Then, based on the risk level of the current scenario (e.g., typhoon evacuation at level 5) and the trainee's experience tag (new employee κ=0.5), the system dynamically adjusts the weighting of each component. Finally, the conformity index for the synthesized action is calculated. in, For the first The specification conformity index of an atomic action The pose dimension score is calculated using the DTW formula. The time-series score is derived from the edit distance formula; The visual dimension score is derived from the fixation percentage formula. Tool usage score (e.g., whether a torque wrench was used, "yes" and "no" are indicated by 1 and 0 respectively); Indicates the weight.

[0043] S302: Dynamically adjust risk weights based on scenarios and personnel.

[0044] First, when the exercise starts, the system reads the scenario type (e.g., "Typhoon Emergency Evacuation" corresponding to risk level R=5) and the trainee's start date from the database, automatically determining their experience level (new / intermediate / advanced). Then, it calculates the dynamic weight of each atomic action according to a formula. In high-risk scenarios, the weight of basic actions is amplified. The calculation formula is: in, This indicates that under risk level R, the action... a i Dynamic weights; The ability coefficient is the highest for personnel with longer service and more proficient skills. The base weight of this action is derived from the knowledge graph. Decide; Indicates the risk level of the scenario (1 = low-risk inspection, 5 = extreme scenarios such as typhoons and fires).

[0045] For example, the capability coefficient is defined as follows: S303: Calculate the overall score under the zero-tolerance mechanism: First, after the exercise, the system summarizes all executed atomic actions. and its dynamic weights Then, the number of serious violations (N) and dangerous actions (such as not wearing a seatbelt, operating with live electrical equipment) during the drill are counted, and each violation is deducted directly from the total score according to the set score (e.g., 50 points). The final comprehensive compliance score is: in, This represents the total number of actions to be performed in the standard action sequence. This is the penalty coefficient; For the first The immediate standardization index of each action; This represents the dynamic weight of the action in scene R; This is an indicator function; it is 1 if the action is actually performed, and 0 otherwise (to avoid scoring poorly for actions that have not been practiced). The total weight of all actions that should be performed; The number of serious violations, such as not wearing a seat belt or operating with live electrical equipment.

[0046] In the above formula, the numerator only counts the actual actions performed, avoiding the "score being dragged down by unpracticed actions"; the denominator is normalized to make scores comparable across different scenarios; the last term... It is a "zero tolerance" mechanism. A single fatal violation in offshore wind power can lead to elimination. The formula directly reflects this management requirement and eliminates the possibility of "getting high scores by luck".

[0047] Step 4: Data-driven continuous improvement and profile generation S401: For trainees: The system automatically retrieves students' past records. All training data is used to calculate the long-term standard execution index for each atomic action, weighted by time decay (higher weight for more recent actions). in, Indicates that the academy students are in action a i The long-term normative capability index; Indicates past past The total number of times this action was practiced within a day. For the first The immediate standardization index of this movement during the second drill; This is the current date and timestamp.

[0048] Then, the average CI value of each operating part is displayed on the 3D fan model using a color temperature map (red < 0.6, yellow 0.6-0.8, green > 0.8). Finally, a personal weakness list (Top 3 low-scoring actions) is generated and pushed to the VR training system homepage, requiring trainees to prioritize completing the targeted modules.

[0049] S402: To the contractor: First, the average specification index of all contractor trainees in the previous period is automatically calculated every set period. and standard deviation .

[0050] Then, generate a dynamic baseline: in, This is the dynamic benchmark for compliance; anything below this benchmark is considered substandard. The total number of contractor trainees who participated in the evaluation that month; For the first Individual average standardization index of students (all movements) (mean).

[0051] Identify contractors below the baseline (approximately the bottom 10%). Ultimately, send a safety warning notice to the contractor, requiring them to pay an improvement bond and mandating a "standardization refinement week" for all their employees until the following month's assessment shows they are no longer in the bottom 10%, otherwise their operating license will be suspended.

[0052] Statistical methods are used to dynamically determine the passing score, avoiding the unfairness of fixed score lines. If the overall level of a group of contractors is high (σ is small), the benchmark is automatically raised to ensure "selection of the best among the best." When offshore wind power clients manage dozens of contractors, this formula provides an objective basis for horizontal comparison, solving the management problem of "all qualified but with varying levels of skill."

[0053] Example 2 This embodiment, based on Embodiment 1, incorporates real wind farm data to drive dynamic scenario evolution. Details are as follows: First, LiDAR and oblique photography equipment are deployed at the target wind farm to perform millimeter-level 3D scanning of the wind turbine towers, submarine cable routes, and substations, ensuring that the geometric error of the virtual environment is less than 5mm.

[0054] Then, a data interface is developed to connect in real time to the wind farm's SCADA system (wind turbine vibration, oil temperature), the marine meteorological station (wind speed, wave height), and the ship's AIS system (traffic dynamics), refreshing environmental parameters every 5 seconds. To handle data noise and latency, Kalman filtering is used for state estimation. in, Represents the discrete time step. Let be the posterior state estimation vector, at time... The optimal estimate after fusion of observations For the prior state prediction vector, based on -1 step state pair Prediction of steps; The Kalman gain matrix has dimensions n×m and dynamically balances the confidence levels of predictions and observations. The observation vector has a dimension of m×1 and consists of actual measurements from SCADA, weather stations, and AIS. The observation matrix, with dimensions m×n, maps the state space to the observation space (e.g., from "wind turbine tilt angle" to "accelerometer reading").

[0055] State vector x It includes heterogeneous data from multiple sources, such as wind speed, wave height, wind turbine tilt angle, and ship position. When network latency exceeds 100ms, the system uses second-order extrapolation prediction compensation. in, The current moment; This refers to network latency. The state value before the delay (e.g., wave height); The first derivative of the state; The second derivative of the state; This is the predicted state after compensation.

[0056] Finally, this assimilated data is input into the virtual environment evolution engine, so that the wind turbines swaying and the waves rising and falling in the VR scene are synchronized with the real world, and trainees experience the same unpredictability as in actual combat during training.

[0057] Example 3 This embodiment provides an operational specification assessment system based on a VR safety training system, applying the method in Embodiment 1 or Embodiment 2, including: 1. Data Acquisition Unit: Connects to the VR safety training system to acquire various data during the trainee's operation in real time, and then transmits the data to the data processing and scoring output unit for processing.

[0058] II. Data Processing and Scoring Output Unit: Using the method of Example 1 or Example 2, the long-term standard execution index and dynamic baseline are finally output and transmitted to the decision-making unit.

[0059] Includes the following modules: The standard knowledge graph construction module is used to atomically decompose safety operating procedures into action-object-constraint triples, and assign differentiated weights to the dependencies between actions based on the severity of the risk consequences. The multi-source data fusion scenario-driven module is used to access wind farm SCADA, marine meteorological and ship dynamic data in real time, and drive the dynamic evolution of the virtual environment through Kalman filtering and delay compensation algorithms; The multimodal behavior acquisition module is used to simultaneously acquire the handle posture trajectory data, eye gaze point data, and action timing data of the trainee's operation; The atomic action-level evaluation module is used to calculate posture deviation through dynamic time warping algorithm, calculate key area gaze rate through indicator function, calculate temporal compliance through edit distance, and integrate them into a specification compliance index; The dynamic weighting and comprehensive scoring module is used to dynamically adjust the action weights based on the scenario risk level and the trainee's experience and ability coefficient, and to calculate a comprehensive standardization score that includes rigid penalty items. The output module is continuously optimized to generate individual capability profile heatmaps, dynamic baselines for contractor groups, and signals that trigger procedural reviews.

[0060] III. Decision-making Unit: The decision-making unit makes decisions on colleges or contractors based on the long-term standard execution index and dynamic baseline, such as the bottom-ranking elimination.

[0061] In summary, this invention addresses the critical deficiency of existing VR safety training methods, which fail to provide atomic-level, multi-dimensional, and process-oriented detailed evaluation of operational procedures. This technical solution achieves a precise breakthrough through three innovations: First, it deconstructs general safety procedures into "action-object-constraint" triplets and constructs a knowledge graph, enabling the system to recognize key details such as "the locking fall arrestor must be heard clicking and gently pulled for verification," rather than simply focusing on whether the action is completed. Second, it simultaneously collects multimodal data such as controller posture, eye movement, and operation timing, utilizing dynamic time warping, indicative functions, and edit distance. Algorithms quantify spatial trajectory accuracy, visual inspection sufficiency, and action sequence rigor, respectively, and then integrate them into a compliance index, achieving a paradigm shift from "results-oriented" to "process diagnosis." Finally, through a comprehensive scoring mechanism that includes rigid penalties, fatal violations and minor deviations are differentiated, forming a traceable, comparable, and enforceable quantitative assessment system. This completely solves the core pain points of traditional training, such as the lack of control over contractor personnel's operational details and insufficient identification of risk chain reactions, making safety capability assessment truly refined, standardized, and executable in the industry.

[0062] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0063] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0064] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0065] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for evaluating operational procedures based on a VR safety training system, characterized in that, Includes the following steps: The safety operating procedures are atomically decomposed into action-object-constraint triples to construct a knowledge graph. Data on the handgrip posture trajectory, eye-tracking fixation points, and action timing of trainees' operations are collected. Posture deviation is calculated using a dynamic time warping algorithm, fixation rate in key areas is calculated using an indicator function, and timing compliance is calculated using edit distance. All of these are then integrated into a compliance index. Calculate the comprehensive normative score, which includes rigid penalty items.

2. The method for evaluating operational procedures based on a VR safety training system according to claim 1, characterized in that, The construction of a knowledge graph includes the following steps: Break down safety procedures into sets of atomic actions. ; Set of atomic actions Import the graph database and define the action relationships and weights between them: in, To standardize knowledge graph objects, they must contain all the information of a directed graph; , represents the set of vertices of a graph, where each vertex... Represents an atomic action; Represents the set of edges of a graph; For two atomic actions that have a dependency relationship, This is a preliminary action. For subsequent actions; The weights are assigned based on the risk consequences classification.

3. The method for evaluating operational procedures based on a VR safety training system according to claim 1, characterized in that, Attitude deviation is calculated as follows: in, The three-dimensional trajectory sequence of the handle operated by the student. ; This is the trajectory sequence of the technician's standard operating procedure. ; , indicating time The three-dimensional coordinates of the handle; This provides a set of all possible timing alignment paths to address the issue of inconsistent student operation speeds with the standard. For the alignment function, the first element of the standard sequence is... Frame mapped to the first of the student sequence frame, Given the Euclidean norm, calculate the straight-line distance between two points in space. The length of the trajectory sequence; The posture score is obtained: in, This represents the spatial deviation tolerance threshold; exceeding this value is considered a serious deviation. This represents the weighting coefficient of the pose dimension in the total score.

4. The method for evaluating operational procedures based on a VR safety training system according to claim 1, characterized in that, The fixation rate in key areas is calculated as follows: in, Indicates the compliance rate of observation; This indicates the total operation time of the action; This is an indicator function that returns either 0 or 1. For a moment The three-dimensional coordinates of the eye's gaze point; A 3D bounding box representing a key region; The minimum fixation duration required by the standard; Visual compliance points were deducted. in, This represents the weighting coefficient of the visual dimension in the total score.

5. The method for evaluating operational procedures based on a VR safety training system according to claim 4, characterized in that, Timing compliance is calculated in the following way: in, This represents the sequence of actions actually performed by the trainee; This represents the sequence of actions required by the specification. Represents the set of all possible sequences of edit operations; The k-th editing operation Indicates the cost of editing operations; Indicates the total length of the edit sequence; Resulting in time series scores: in, Indicates the length of the standard action sequence; This represents the weighting coefficient of the time series dimension in the total score.

6. The method for evaluating operational procedures based on a VR safety training system according to claim 5, characterized in that, The compliance index is calculated using the following formula: in, For the first The specification conformity index of an atomic action. Score the pose dimension; Scoring is based on the time-series dimension; Score the visual dimension; Tool usage score; Indicates the weight.

7. The method for evaluating operational procedures based on a VR safety training system according to claim 6, characterized in that, The comprehensive standard score is calculated as follows: in, This represents the total number of actions to be performed in the standard action sequence. This is the penalty coefficient; For the first The immediate standardization index of each action; This represents the dynamic weight of the action in scene R; This is an indicator function; it is 1 if the action is actually performed, and 0 otherwise. The total weight of all actions that should be performed; This refers to the number of serious violations.

8. An operational procedure evaluation system based on a VR safety training system, characterized in that, The method applied to any one of claims 1 to 7 includes: Data acquisition unit: Connects to the VR safety training system to acquire various data during the trainee's operation in real time, and then transmits the data to the data processing and scoring output unit for processing; Data processing and scoring output unit: Using the method of Example 1 or Example 2, the long-term compliance index and dynamic baseline are finally output, and the long-term compliance index and dynamic baseline are transmitted to the decision-making unit; Decision-making unit: The decision-making unit makes decisions on colleges or contractors based on long-term standard implementation index and dynamic baseline.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the operation specification evaluation method of the VR-based safety training system according to any one of claims 1 to 7.

10. A processor, characterized in that, The processor is used to run a program, wherein the program executes the operation specification evaluation method of the VR-based safety training system according to any one of claims 1 to 7 when it runs.

Images