Capability migration prediction method and system in industrial skill practical training

Abstract

A capability migration prediction method and system in industrial skill practical training relates to the field of occupational skill assessment, and the method comprises the following steps: firstly, obtaining a constraint condition type of virtual environment relative to real equipment simplification, and recording an operation performance difference value as constraint condition dependence intensity by setting a constraint-containing and unconstrained control group and experiment group virtual environment; then, analyzing constraint conditions related to each link of a real machine task, calculating single capability loss based on dependency intensity and a virtual-real difference multiple, and performing weighted summation by considering coupling coefficients among the constraint conditions to obtain a total capability loss predicted value; and finally, the predicted value is converted into real machine expected data, and a virtual-real ability migration evaluation result of the target student is generated by comparing the real machine expected data with actual data. By implementing the method, the refinement degree of virtual and real capability migration evaluation can be improved.

Description

Technical Field

[0001] This application relates to the field of vocational skills assessment, and in particular to a method and system for predicting ability transfer in industrial skills training. Background Technology

[0002] With the deepening of vocational education informatization, virtual training systems have been widely used in the field of skills training. Virtual training environments simulate the operation scenarios of real equipment using computer simulation technology, providing students with opportunities for repeated practice and reducing training costs and safety risks. However, there are objective differences between virtual environments and real equipment; a student's good performance in a virtual environment does not guarantee their competence in real-world operation.

[0003] In related technologies, the effectiveness of skill transfer can be evaluated by comparing trainees' performance in virtual and real-device environments. Specifically, a real-device assessment is conducted after virtual training, recording trainees' actual performance on the real machine. The real-device scores are then compared and analyzed with the virtual training scores to determine the effectiveness of the virtual training. Some systems will also analyze typical error types encountered by trainees in the real-device environment during the comparison process, categorizing them into macro-level categories such as lack of operational proficiency and poor environmental adaptability, serving as a reference for subsequent training improvements.

[0004] However, with the increasing demand for refined training management, the above-mentioned solution has revealed significant limitations in practical application. The assessment conclusions obtained through overall performance comparison are rather crude, attributing trainees' errors on real-world machines to macro-level categories such as unfamiliarity with operation or poor environmental adaptability. It fails to specify the degree of skill decline experienced by trainees in particular operational stages, nor can it explain the differences in real-world performance among different trainees with the same virtual training scores. This results in insufficient relevance of the assessment results for subsequent training guidance, reducing the practical value of skills transfer assessment. Summary of the Invention

[0005] This application provides a method and system for predicting ability transfer in industrial skills training, which can improve the precision of virtual-real ability transfer assessment.

[0006] Firstly, this application provides a method for predicting capability transfer in industrial skills training, applied to a capability transfer prediction system. The method includes: acquiring simplified constraint types in a virtual environment relative to real equipment, including physical space constraints, information integrity constraints, operational feedback delay constraints, and error tolerance constraints; for each constraint type, setting up a control virtual environment containing the constraint type and an experimental virtual environment without the constraint type, recording the difference in operational performance indicators between the control and experimental virtual environments, and using the difference in operational performance indicators as the target student's dependence on the corresponding constraint type; extracting the constraint types involved in each operational step of the real machine task, determining individual capability loss values ​​based on the dependence strength corresponding to each constraint type and the difference multiple between the real machine and the virtual environment, calculating the coupling coefficient between different constraint types within the operational step, and weighted summing all individual capability loss values ​​with their corresponding coupling coefficients to obtain the total capability loss prediction value for the operational step; converting the total capability loss prediction value into expected data for the real machine task, collecting actual data of the target student in the real machine environment, and generating a virtual-to-real capability transfer evaluation result value for the target student based on the deviation between the expected and actual data.

[0007] In the above embodiments, four simplified constraint types of the virtual environment are obtained. By setting up a control environment and an experimental environment, the dependence strength of trainees on each constraint is measured. The constraint types involved in each operation of the real machine task are extracted and the coupling coefficient is calculated. Based on the dependence strength, difference multiple and coupling relationship, the predicted value of total capability loss is determined. The predicted value is converted into expected data and compared with the actual data of the real machine to generate the virtual-real capability transfer assessment result value. This solves the problem that the traditional solution attributes real machine failures to macro categories and cannot specifically explain the extent of capability decline of trainees in which operation stages, thus improving the precision of virtual-real capability transfer assessment.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of extracting the constraint types involved in each operation step in the real machine task, and determining the individual capability loss value based on the dependency strength corresponding to each constraint type and the difference multiple between the real machine and the virtual environment, specifically includes: obtaining the operation flow of the real machine task and dividing the operation flow into multiple operation steps; for each operation step, analyzing one or more constraint types among the physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints involved in the operation step; obtaining the reference benchmark value of each constraint type in the real machine environment and the corresponding actual value in the virtual environment; calculating the difference multiple of each constraint type based on the ratio of the reference benchmark value to the actual value; calculating the preliminary loss value based on the dependency strength and the difference multiple; and normalizing the preliminary loss value based on the standard completion time of the operation step to obtain the individual capability loss value.

[0009] In the above embodiments, the operation steps of the real machine task are divided and the types of constraints involved in each step are analyzed. The parameter benchmark values ​​of the real machine environment and the virtual environment are obtained to calculate the difference multiple. After multiplying the dependency strength by the difference multiple to obtain the preliminary loss value, the loss value of each capability is obtained by normalization according to the standard completion time of the operation step. This makes the loss values ​​of operation steps of different durations comparable and provides a unified quantitative standard for subsequent cross-step capability loss comparison analysis.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the step of calculating the coupling coefficient between different constraint types within the operation phase and weighted summing all individual capability loss values ​​with their corresponding coupling coefficients to obtain the predicted total capability loss value of the operation phase specifically includes: for each pair of constraint types within the operation phase, recording the first operational performance index of the target trainee in an experimental virtual environment where two constraint types are removed simultaneously; comparing the first operational performance index with the second operational performance index when removing individual constraint types separately, and calculating the coupling coefficient between the two constraint types; constructing a coupling matrix containing all constraint types within the operation phase; and weighted summing all individual capability loss values ​​within the operation phase based on the coupling coefficients in the coupling matrix to obtain the predicted total capability loss value.

[0011] In the above embodiment, the operational performance index of the trainee when removing two constraints at the same time is recorded. The coupling coefficient is calculated by comparing the index with the performance index when removing each constraint individually. A coupling matrix containing all constraint types is constructed. The total capability loss prediction value is obtained by weighted summation of individual capability loss values ​​based on the coupling matrix. This captures the synergistic or interference effects between different constraint types and avoids the prediction bias caused by simply superimposing the effects of each constraint.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of generating the virtual-to-real ability transfer assessment result value of the target trainee based on the deviation data between expected data and actual data specifically includes: calculating the mean and standard deviation of the total ability loss prediction value of each operation link, and marking the operation links that exceed the mean by a preset multiple of the standard deviation as the links to be analyzed; for the links to be analyzed, calculating the deviation ratio between expected data and actual data, and identifying the operation links with the largest number of constraint types and deviation ratios exceeding a preset threshold as the weak links; extracting the individual ability loss value corresponding to each constraint type in the weak links, sorting the individual ability loss values ​​according to their numerical values ​​and normalizing them to obtain the degree of influence of the constraint type; and generating the virtual-to-real ability transfer assessment result value of the target trainee based on the distribution location of the weak links and the degree of influence of the constraint types.

[0013] In the above embodiments, the system calculates the mean and standard deviation of the total capability loss prediction value to mark the links to be analyzed, determines the weak links by the ratio of the expected data to the actual data of the links to be analyzed, extracts the individual capability loss values ​​of each constraint type in the weak links and normalizes them to obtain the degree of influence, and generates evaluation result values ​​based on the distribution location of the weak links and the degree of influence of the constraints, thus locating the capability loss to the specific operation link and the dominant constraint type, providing a basis for formulating targeted reinforcement training programs.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the step of generating the virtual-to-real capability transfer assessment result value of the target trainee based on the distribution location of capability weaknesses and the degree of influence of constraint types specifically includes: calculating the distribution density of capability weaknesses in all operational stages, and marking continuously distributed capability weaknesses as key attention intervals; calculating the cumulative impact value of each constraint type within the key attention interval according to the degree of influence of constraint types; generating interval capability loss coefficients based on the weighted combination of the distribution density and cumulative impact values ​​of the key attention intervals; and weighting and summing the interval capability loss coefficients with the total capability loss prediction value to obtain the virtual-to-real capability transfer assessment result value.

[0015] In the above embodiments, the distribution density of the weak links in the system's computing power is marked as the key focus interval. The cumulative impact value of each constraint within the interval is calculated according to the degree of influence of the constraint. An interval capability loss coefficient is generated based on the distribution density and the cumulative impact value. The interval coefficient is weighted and summed with the total capability loss prediction value to obtain the evaluation result value. This identifies the concentrated area of ​​the student's capability transfer problem and reflects the impact of the cumulative effect of constraints on capability transfer in the continuous operation process.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of generating the virtual-to-real ability transfer assessment result value of the target learner based on the deviation data between expected data and actual data, the method further includes: constructing a virtual training scenario library, setting multiple difficulty levels of training scenarios for each constraint type; matching training scenarios of corresponding difficulty levels to the target learner based on the virtual-to-real ability transfer assessment result value; setting a progressive constraint adjustment mechanism in the training scenario, the constraint adjustment mechanism including gradually increasing physical space restrictions, reducing operation prompt information, shortening operation feedback time, and reducing error tolerance; recording the adaptation process of the target learner in scenarios of different difficulty levels, and generating a personalized training advancement path.

[0017] In the above embodiments, the system constructs a virtual training scenario library and sets up scenarios with multiple difficulty levels. Based on the evaluation results, it matches the corresponding difficulty scenario to the trainee. In the scenario, a progressive adjustment mechanism is set up to gradually increase physical space restrictions, reduce operation prompts, shorten feedback time and reduce error tolerance. The system records the trainee's adaptation process in different difficulty scenarios to generate a personalized training advancement path, realizing a smooth transition from the virtual environment to the real machine environment and shortening the trainee's adaptation period from virtual training to real machine operation.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, the step of constructing a virtual training scenario library and setting multiple difficulty levels for training scenarios for each constraint type specifically includes: dividing the training scenarios into basic action training scenarios, single skill training scenarios, and comprehensive skill training scenarios according to their operational nature; setting physical space constraints and operation feedback delay constraints in basic action training scenarios; setting information integrity constraints and error tolerance constraints in single skill training scenarios; and setting all constraint types in comprehensive skill training scenarios, and dynamically adjusting the parameters of each constraint type according to the training performance of the target learner.

[0019] In the above embodiments, the system divides the training scenarios into three categories: basic actions, single skills, and comprehensive skills. Physical space constraints and operation feedback delay constraints are set in the basic action scenarios, information integrity constraints and error tolerance constraints are set in the single skill scenarios, and all constraints are set in the comprehensive skill scenarios. The parameters are dynamically adjusted according to the trainee's performance. Corresponding combination of constraints is matched for different operation natures, forming a progressive training system from single actions to complex tasks.

[0020] In a second aspect, embodiments of this application provide a capability migration prediction system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the capability migration prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a capability migration prediction system, cause the capability migration prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a capability migration prediction system, cause the capability migration prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the capability migration prediction system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application obtains four simplified constraint types from the virtual environment, measures the trainee's dependence on each constraint by setting up a control environment and an experimental environment, extracts the constraint types involved in each operation of the real machine task and calculates the coupling coefficient, determines the predicted value of total capability loss based on dependence strength, difference factor and coupling relationship, converts the predicted value into expected data and compares it with the actual data of the real machine to generate the virtual-real capability transfer assessment result value. This solves the problem of traditional solutions attributing real machine failures to macro categories and failing to specify the extent of capability decline in which operation stages the trainee will experience, thus improving the precision of virtual-real capability transfer assessment.

[0026] 2. This application divides the operation steps of the real machine task and analyzes the types of constraints involved in each step. It obtains the parameter benchmark values ​​of the real machine environment and the virtual environment, calculates the difference multiple, multiplies the dependency intensity by the difference multiple to obtain the preliminary loss value, and then normalizes it according to the standard completion time of the operation step to obtain the individual capability loss value. This makes the loss values ​​of operation steps of different durations comparable and provides a unified quantitative standard for subsequent cross-step capability loss comparison analysis.

[0027] 3. This application records the operational performance index of trainees when removing two constraints simultaneously, compares this index with the performance index when removing each constraint individually to calculate the coupling coefficient, constructs a coupling matrix containing all constraint types, and obtains the predicted value of total capability loss by weighted summation of individual capability loss values ​​based on the coupling matrix. This captures the synergistic or interference effects between different constraint types and avoids the prediction bias caused by simply superimposing the effects of each constraint. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a capability transfer prediction method in industrial skills training as described in this application.

[0029] Figure 2 This is another flowchart illustrating the ability transfer prediction method in industrial skills training as described in this application.

[0030] Figure 3 This is a schematic diagram of the physical device structure of a capability migration prediction system in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] This application's embodiments are primarily applied in the fields of vocational education and industrial skills training, particularly in scenarios where there is a need for skill transfer between virtual training and real machine operation. For example, in CNC machine tool operation skills training, trainees first undergo operational training in a virtual simulation system. The virtual environment simplifies physical space limitations, provides complete operational prompts, and allows trainees to repeatedly try and fail. When trainees need to transition to real CNC machine tool operation after completing virtual training, the system uses the method described in this application to measure the trainee's dependence on various simplified constraints in the virtual environment, predicting potential skill losses in the real machine environment, such as clamping errors due to physical space constraints, program editing errors due to a lack of operational prompts, and overtravel risks due to delayed operational feedback. This provides teachers with targeted reinforcement training suggestions, helping trainees better transfer their skills from virtual training to real machine operation.

[0035] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a capability transfer prediction method in industrial skills training as described in this application.

[0036] S101. Obtain the simplified constraint types of the virtual environment relative to the real device. The constraint types include physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints.

[0037] In this context, "virtual environment" refers to a training system that simulates real-world equipment operation scenarios using computer simulation technology. Real-world equipment refers to the physical equipment used in actual production or teaching. "Constraint type" refers to the simplified classification of operational limitations in a virtual environment compared to real-world equipment. Physical space constraints refer to physical geometric limitations imposed during operation, such as spatial dimensions, work area, operator posture, and tool placement. For example, arm extension is limited in real machine tool operation, but this limitation may be eliminated in a virtual environment. Information integrity constraints refer to the degree of completeness of information the operator obtains regarding equipment status, process parameters, fault prompts, and operating instructions. For example, real equipment may only display some parameters through instruments, while a virtual environment may provide a visual interface for all parameters. Operation feedback delay constraints refer to the time delay between executing an operation and the system's response. For example, mechanical inertia in real hydraulic systems causes response delays, while a virtual environment may provide instantaneous responses. Error tolerance constraints refer to the system's tolerance for operational errors and the limitations on the opportunities for error correction. For example, misoperation of real equipment may lead to equipment damage, while a virtual environment typically allows for undoing or restarting operations.

[0038] Specifically, the system first conducts a comprehensive analysis of the operating environment of the real equipment, identifying various simplified constraints in the virtual simulation system's design to reduce training difficulty, development costs, or ensure training safety. The system categorizes these simplified constraints into four types based on their mechanism of action and impact: physical space constraints, information integrity constraints, operational feedback delay constraints, and error tolerance constraints. The system records the specific manifestation, parameter range, and degree of restriction of each constraint type in the real equipment, serving as a benchmark for subsequent dependency strength measurement and capability loss prediction.

[0039] In some embodiments, constraint types can be obtained in multiple ways. Optionally, constraint types can be identified through comparative analysis, specifically including collecting the operating space dimensions, information display interface parameters, system response time data, and error handling mechanism rules of the real equipment, simultaneously collecting the corresponding simulation parameters and settings in the virtual environment, and comparing and analyzing them item by item to identify and classify constraint dimensions with significant differences. Optionally, constraint types can be determined through expert evaluation, specifically including organizing technical experts and senior operators in the field of equipment operation to conduct experience testing of the virtual simulation system. Experts will point out specific constraints in the virtual environment that are simplified compared to the real equipment based on their actual operating experience. The system will then summarize the expert opinions into four constraint types and record the typical characteristics of each type. It is understood that other methods can also be used to obtain constraint types, which are not limited here.

[0040] S102. For each constraint type, set up a control virtual environment containing the constraint type and an experimental virtual environment removing the constraint type. Record the difference in operational performance indicators between the target student in the control virtual environment and the experimental virtual environment. Use the difference in operational performance indicators as the target student's dependence on the corresponding constraint type.

[0041] The control virtual environment represents a virtual simulation scenario that retains specific constraint types, while the experimental virtual environment represents a virtual simulation scenario where the restrictions of specific constraint types have been removed or reduced. Operational performance indicators are quantitative values ​​that measure the quality of a student's operation, including dimensions such as completion time, accuracy, and number of errors. The difference in operational performance indicators refers to the numerical difference in operational performance indicators between the control and experimental virtual environments. Dependency strength represents the degree to which a student's operational ability depends on a specific constraint type; a higher value indicates a more significant decrease in student operational performance after the constraint is removed.

[0042] Specifically, the system constructs two versions of virtual environments for each type of constraint. The control virtual environment maintains the original settings for that constraint type, while the experimental virtual environment removes or significantly reduces the restriction level of that constraint type. The system assigns the target learner to perform the same operational task in both environments and records the learner's operational performance indicators in both environments. The system calculates the difference between the operational performance indicators in the control and experimental virtual environments; this difference reflects the magnitude of change in the learner's operational ability after the constraint is removed. The system stores this difference as the strength of the target learner's dependence on the corresponding constraint type.

[0043] In some embodiments, dependency strength can be measured in various ways. Optionally, a single-task testing method can be used, specifically including designing a standardized operational task for a specific constraint type, recording the time and number of errors for trainees to complete the task in a control virtual environment, repeating the same task in an experimental virtual environment and recording the corresponding indicators, and calculating the difference between the two test results as the dependency strength. Optionally, a multi-task comprehensive testing method can be used, specifically including designing a comprehensive task covering multiple operational stages, testing it separately in a control virtual environment and an experimental virtual environment, and weighted averaging the differences in operational performance indicators for each stage to obtain the comprehensive dependency strength for that constraint type. It is understood that other methods can also be used to measure dependency strength, and this is not limited here.

[0044] S103. Extract the constraint types involved in each operation step in the real machine task, determine the individual capability loss value based on the dependency strength corresponding to each constraint type and the difference multiple between the real machine and the virtual environment, calculate the coupling coefficient between different constraint types in the operation step, and obtain the total capability loss prediction value of the operation step by weighted summing of all individual capability loss values ​​and their corresponding coupling coefficients.

[0045] In this context, "operational stage" refers to an independent operational step or phase within the real-device task operation process. "Dependency strength" represents the degree to which a learner depends on a specific constraint type. "Difference ratio" refers to the ratio of parameters between the real-device environment and the virtual environment for a specific constraint type. "Individual capability loss value" represents the degree of capability decline experienced by a learner when migrating from a virtual environment to a real-device environment under the influence of a single constraint type. "Coupling coefficient" refers to the correlation strength between different constraint types. "Total capability loss prediction value" represents the predicted overall capability loss of a learner in a specific operational stage after comprehensively considering all constraint types and their coupling relationships.

[0046] Specifically, the system first decomposes the complete operation process of the real-device task into multiple independent operation steps, and analyzes the types of constraints involved in each operation step. The system obtains the dependency strength value corresponding to each type of constraint involved in the operation step, and simultaneously calculates the difference factor between the real-device environment and the virtual environment for that constraint type. The system multiplies the dependency strength by the difference factor to obtain a preliminary loss value, and then normalizes it according to the standard completion time of the operation step to obtain the individual capability loss value. The system measures the interaction effect between different constraint types through controlled variable experiments, constructs a coupling matrix to record the coupling coefficient between each pair of constraint types. The system then performs a weighted summation of all individual capability loss values ​​within the operation step with their corresponding coupling coefficients to obtain the predicted total capability loss value for that operation step.

[0047] In some embodiments, the total capacity loss prediction can be calculated in several ways. Optionally, a linear weighted method can be used, specifically involving multiplying each individual capacity loss value by its corresponding coupling coefficient and then summing the results. The coupling coefficient is extracted from the coupling matrix. For components involving multiple constraint types, matrix operations are used to obtain a comprehensive weight, and the weighted sum is used as the total capacity loss prediction. Optionally, a nonlinear mapping method can be used, specifically involving first calculating the weighted sum of each individual capacity loss value as the base loss, then introducing a nonlinear correction factor based on the number of constraint types and the distribution characteristics of the coupling coefficients, and adjusting the base loss using an exponential or logarithmic function to obtain a more realistic total capacity loss prediction. It is understood that other methods can also be used to calculate the total capacity loss prediction, and this is not limited here.

[0048] S104. Convert the total capability loss prediction value into expected data for real machine tasks, collect the actual data of the target trainees in the real machine environment, and generate the virtual-real capability transfer assessment result value of the target trainees based on the deviation data between the expected data and the actual data.

[0049] Expected data refers to the student's anticipated performance in a real-world environment, calculated based on the predicted total ability loss. Actual data represents the objective operational data recorded when the student actually completes the task in the real-world environment. Deviation data refers to the quantified difference between expected and actual data. The virtual-to-real ability transfer assessment result is used to represent a comprehensive score of the student's ability retention when migrating from a virtual training environment to a real-world operating environment.

[0050] Specifically, based on the predicted total capability loss for each operational step and benchmark data of the learner's performance in the virtual environment, the system calculates the expected completion time, expected error rate, and other expected data for each operational step in the real-world environment. When the learner actually performs real-world operations, the system collects actual data such as the learner's actual completion time and the number of actual errors through sensors, timers, and operation logs. The system calculates the deviation between the expected and actual data, marking and deeply analyzing operational steps that exceed a preset deviation threshold. The system integrates multi-dimensional information such as the deviation distribution of each operational step, the location of weak points, and the degree of influence of constraint types, and generates a weighted calculation result value for the target learner's virtual-to-real-world capability transfer assessment.

[0051] In some embodiments, the virtual-to-real capability migration assessment result value can be generated in multiple ways. Optionally, a deviation statistics method can be used, specifically including calculating the absolute value of the deviation between expected and actual data for all operational steps, weighting and summing the deviation values ​​according to the importance of the operational steps, and mapping the weighted sum to a scoring range of 0 to 100 through normalization to obtain the virtual-to-real capability migration assessment result value. Optionally, a hierarchical assessment method can be used, specifically including first identifying weak links in capability and calculating their distribution density in the overall process, then extracting the types of dominant constraints involved in each weak link and their degree of influence, and finally combining the distribution density, degree of influence, and predicted total capability loss value through multi-level weighted calculation to obtain the virtual-to-real capability migration assessment result value. It is understood that other methods can also be used to generate the virtual-to-real capability migration assessment result value, which are not limited here.

[0052] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the ability transfer prediction method in industrial skills training in this application embodiment.

[0053] S201. Obtain the simplified constraint types of the virtual environment relative to the real device. The constraint types include physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints.

[0054] Physical space constraints refer to the physical and geometric limitations imposed on operation, such as spatial dimensions, working range, and posture angles. For example, the size of a real machine tool's worktable is 800mm × 600mm, while in a virtual environment it is set to infinitely large. Information integrity constraints refer to the limitations on the completeness of information obtained regarding equipment status, process parameters, and fault information. For example, a real machine may only display 3 key parameters, while a virtual environment displays all 15. Operation feedback latency constraints refer to the time delay between executing an operation and the system's response. For example, a real hydraulic system may have a response delay of 200ms, while in a virtual environment it is 10ms. Error tolerance constraints refer to the system's tolerance for operational errors. For example, a real machine may suffer permanent damage due to misoperation, while a virtual environment allows for an unlimited number of retries.

[0055] The system identifies constraint types through parameter comparison. For physical space constraints, it measures the operating space dimensions of the real device and compares them with the virtual environment's settings. For information integrity constraints, it counts the number of information items displayed on the real device and the number of information items provided by the virtual environment. For operation feedback latency constraints, it measures the time from command issuance to execution completion on the real device and the corresponding time in the virtual environment. For error tolerance constraints, it records the number of errors allowed on the real device and the error tolerance settings of the virtual environment.

[0056] S202. For each constraint type, set up a control virtual environment containing the constraint type and an experimental virtual environment removing the constraint type. Record the difference in operational performance indicators between the target student in the control virtual environment and the experimental virtual environment. Use the difference in operational performance indicators as the target student's dependence on the corresponding constraint type.

[0057] The control virtual environment refers to a virtual scene with specific constraints, such as setting the spatial boundary to 800mm × 600mm. The experimental virtual environment refers to a virtual scene without these constraints, such as removing the spatial boundary restrictions. Operational performance indicators include quantitative values ​​such as completion time, number of errors, and operational accuracy. Dependency strength refers to the degree to which the trainee's operational ability depends on specific constraints, and its value is determined by the difference between the performance indicators in the two environments.

[0058] To address physical space constraints, the system constructs a control environment with spatial limitations preserved and an experimental environment without spatial limitations. Trainees complete the same tasks in both environments. The completion time in the control environment is recorded as 120 seconds with 3 errors, while the completion time in the experimental environment is recorded as 80 seconds with 1 error. The time difference of 40 seconds and the error difference of 2 are calculated, and the normalized difference is taken as the trainee's dependence on the physical space constraints. This process is repeated for the other three types of constraints to obtain the dependence values ​​on information integrity, operation feedback delay, and error tolerance, respectively.

[0059] S203, Obtain the operation process of the real machine task, and divide the operation process into multiple operation steps.

[0060] Among them, "real machine task" refers to the complete operational task that trainees need to complete on real equipment, such as a CNC machine tool parts machining task. "Operation process" refers to the sequence of operational steps required to complete the task. "Operation link" refers to the independent step or stage in the operation process, such as workpiece clamping, tool selection, program editing, machining execution, and quality inspection.

[0061] The system acquires the standard operating procedure document for a real machine task, which details all the steps to complete the task. Based on the functional attributes and time nodes of each step, the system divides the operation process into multiple independent operation stages. For example, a CNC machining task might be divided into five stages: Stage 1 is workpiece clamping, Stage 2 is tool installation and tool setting, Stage 3 is machining program editing, Stage 4 is parameter setting and trial run, and Stage 5 is formal machining and monitoring. The system assigns a unique identifier to each operation stage and records its position number within the overall process, establishing an operation stage database to store the name, estimated completion time, operation requirements, and other attribute information of each stage.

[0062] S204. For each operation step, analyze one or more of the following constraint types: physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints.

[0063] The constraint types involved refer to the specific categories of constraints and restrictions imposed on a particular operational step during execution. An operational step may be affected by a single constraint type or multiple constraint types simultaneously. For example, the workpiece clamping step mainly involves physical space constraints and operational feedback delay constraints, while the program editing step mainly involves information integrity constraints and error tolerance constraints.

[0064] The system performs constraint analysis on each operation step. For the workpiece clamping step, it analyzes whether the operating space is limited to determine the physical space constraints and analyzes the clamping force feedback delay to determine the operation feedback time delay constraints. For the program editing step, it analyzes the required parameter information to determine the information integrity constraints and analyzes the program error modifiability to determine the error tolerance constraints. The system establishes a constraint correlation matrix, where rows represent operation steps and columns represent four types of constraint conditions. A matrix element of 1 indicates that the step involves that constraint, and 0 indicates that it does not. For example, the correlation vector for step 1 is [1, 0, 1, 0], indicating that it involves both physical space constraints and operation feedback time delay constraints.

[0065] S205. Obtain the reference baseline value for each constraint type in the real machine environment and the corresponding actual value in the virtual environment.

[0066] The reference value refers to the actual parameter value of the constraint type in the real machine environment, such as an operating space of 800mm or 3 information display items in the real equipment. The actual value refers to the simulation parameter value of the corresponding constraint in the virtual environment, such as an operating space of 2000mm or 15 information display items in the virtual environment. The difference between the reference value and the actual value reflects the degree of simplification of the virtual environment compared to the real machine environment.

[0067] The system acquires parameters of the real machine environment through on-site measurements. For physical space constraints, measuring tools are used to obtain physical dimensions such as the size of the operating table and the space for tool placement, which are recorded as reference values. For information integrity constraints, the number of information items available on the real machine's display screen is counted. For operation feedback delay constraints, a high-precision timer is used to measure the time interval from the issuance of the operation command to the completion of the actuator's action. For error tolerance constraints, the maximum number of errors allowed by the real machine or the parameters of the recovery mechanism after an error are recorded. Simultaneously, the system reads the configuration file of the virtual environment, extracts the simulation parameters of the corresponding constraints as actual values, and establishes a parameter lookup table to store the reference values ​​and actual values ​​for each type of constraint.

[0068] S206. Calculate the difference multiple for each constraint type based on the ratio of the reference value to the actual value.

[0069] The difference factor refers to the ratio of the reference value in the real-device environment to the actual value in the virtual environment, used to quantify the degree of simplification of the virtual environment. A larger difference factor indicates a higher degree of simplification of the constraint in the virtual environment, and a greater difficulty in adaptation for learners when migrating from the virtual environment to the real-device environment.

[0070] The system calculates the difference factor for each constraint type. For physical space constraints, if the actual machine's operating space is 800mm and the virtual environment's is 2000mm, then the difference factor = 800 / 2000 = 0.4. For information integrity constraints, if the actual machine displays 3 parameters and the virtual environment displays 15 parameters, then the difference factor = 3 / 15 = 0.2. For operation feedback delay constraints, if the actual machine's response delay is 200ms and the virtual environment's delay is 10ms, then the difference factor = 200 / 10 = 20. For error tolerance constraints, if the actual machine allows 0 errors and the virtual environment allows 5 errors, then the difference factor = 0 / 5 = 0 (setting a lower limit of 0.1 avoids division by zero). The system stores the calculated difference factors in the constraint parameter table, providing a quantitative basis for subsequent capacity loss value calculations.

[0071] S207. The preliminary loss value is calculated based on the dependence strength and the difference factor.

[0072] Here, dependency strength refers to the numerical value of a learner's dependence on a specific type of constraint, which has been measured in S202. The difference factor refers to the ratio of parameters between the real machine and the virtual environment under that constraint, which has been calculated in S206. The preliminary loss value is the estimated ability reduction calculated by combining dependency strength and difference factor, and has not yet been normalized.

[0073] The system calculates the initial loss value for each type of constraint involved in each operation. The calculation formula is: Initial Loss Value = Dependency Strength × |1 - Difference Multiple|. For example, if a student's dependency strength on the physical space constraint is 0.6 and the difference multiple for that constraint is 0.4, then the initial loss value = 0.6 × |1 - 0.4| = 0.6 × 0.6 = 0.36. If an operation involves multiple constraints, the initial loss value for each constraint is calculated separately. For example, in the workpiece clamping stage, which involves both physical space constraints and operation feedback delay constraints, the initial loss value for the physical space constraint is 0.36, and the dependency strength and difference multiple for the operation feedback delay constraint are 0.5 and 20, respectively. The initial loss value = 0.5 × |1 - 20| = 0.5 × 19 = 9.5. The system stores the initial loss values ​​for each constraint in the loss value calculation table.

[0074] S208. Normalize the initial loss value according to the standard completion time of the operation to obtain the individual capability loss value.

[0075] The standard completion time refers to the expected completion time of an operation under normal circumstances; for example, the standard completion time for workpiece clamping is 180 seconds. Normalization involves adjusting the initial loss value proportionally to the standard completion time, making the loss values ​​of operations with different durations comparable. The individual capability loss value is the normalized capability loss value, reflecting the degree of impact of a single constraint type on that operation.

[0076] The system obtains the standard completion time for each operation step, extracted from operation procedure documents or historical data. The system normalizes the initial loss value using the formula: Individual Capability Loss Value = Initial Loss Value / Standard Completion Time × 100. For example, the initial loss value for the physical space constraint of the workpiece clamping step is 0.36, and the standard completion time is 180 seconds, then the individual capability loss value = 0.36 / 180 × 100 = 0.2. The initial loss value for the operation feedback delay constraint is 9.5, and the individual capability loss value = 9.5 / 180 × 100 = 5.28. Normalization allows for direct comparison of capability loss values ​​for operation steps of different durations. The system stores the calculated individual capability loss values ​​in the capability loss database, providing a data foundation for subsequent total capability loss prediction.

[0077] S209. For each pair of constraint types in the operation process, record the first operational performance index of the target trainee in the experimental virtual environment where two constraint types are removed simultaneously.

[0078] Each pair of constraint types refers to any combination of two different constraint types involved in the operation, such as physical space constraints and operation feedback delay constraints forming a pair. Simultaneously removing two constraint types in an experimental virtual environment means removing both constraints in the virtual scene at the same time, such as removing both spatial boundaries and feedback delay. The first operational performance index refers to the performance value of the trainee in completing the operation task in this experimental environment, including completion time, number of errors, etc.

[0079] The system identifies all constraint types involved in the operation and generates all possible pairwise combinations. For example, if a certain step involves three types of constraints: physical space constraints, information integrity constraints, and operation feedback delay constraints, then three pairs of combinations are generated. For the first pair of combinations (physical space constraints + information integrity constraints), the system constructs a virtual environment that simultaneously cancels these two constraints, arranges for the trainee to complete a standard operation task, and records the completion time as 65 seconds and the number of errors as 0. For the second pair of combinations (physical space constraints + operation feedback delay constraints), the corresponding environment is constructed and the performance indicators are recorded. This process is repeated for the third pair of combinations, and the first operation performance indicators corresponding to all combinations are stored in the coupling test data table.

[0080] S210. Compare the first operational performance index with the second operational performance index when removing individual constraint types respectively, and calculate the coupling coefficient between the two constraint types.

[0081] The second operational performance index refers to the trainee's operational performance value when removing only a single constraint type; this value is recorded in S202. The coupling coefficient is the strength coefficient of the interaction between two constraint types, reflecting the synergistic or disruptive effects produced when both constraints are removed simultaneously. A coupling coefficient greater than 1 indicates a positive synergistic effect between the two constraints, while a coefficient less than 1 indicates a negative disruptive effect.

[0082] The system extracts the first operational performance index and its corresponding two second operational performance indices for calculation. For example, the completion time is 65 seconds when both physical space constraints and information integrity constraints are removed simultaneously, 80 seconds when only physical space constraints are removed, and 90 seconds when only information integrity constraints are removed. The theoretical superposition effect is calculated as follows: if the two constraints act independently, the expected completion time is (80 + 90) / 2 = 85 seconds. The ratio of the actual effect to the theoretical effect is calculated: the coupling coefficient = 85 / 65 = 1.31. This coefficient greater than 1 indicates that removing both constraints simultaneously produces a positive synergistic effect. The system repeats this calculation process for all constraint type combinations and stores the resulting coupling coefficients in the coupling relationship database.

[0083] S211. Construct a coupling matrix that includes all constraint types within the operational process.

[0084] The coupling matrix is ​​a matrix structure that records the pairwise coupling relationships between all constraint types within an operational process. The rows and columns of the matrix represent different constraint types, and the matrix elements represent the coupling coefficients between corresponding constraint types. The diagonal elements of the matrix are all 1s, representing the coupling coefficient between the constraint and itself. The matrix is ​​symmetric because the coupling relationship between two constraints is mutual.

[0085] The system determines the matrix dimension based on the number and types of constraints involved in the operational steps. For example, if a step involves three types of constraints: physical space constraints, information integrity constraints, and operational feedback delay constraints, a 3×3 matrix is ​​constructed. The system fills the matrix elements: diagonal elements are set to 1, and off-diagonal elements are filled with coupling coefficients calculated in S210. For example, the element in the 1st row and 2nd column is the coupling coefficient of 1.31 for physical space constraints and information integrity constraints; the element in the 1st row and 3rd column is the coupling coefficient of 0.87 for physical space constraints and operational feedback delay constraints; and the element in the 2nd row and 3rd column is the coupling coefficient of 1.15 for information integrity constraints and operational feedback delay constraints. The system fills the lower triangular part of the matrix according to symmetry, completing the construction and storage of the coupling matrix.

[0086] S212. Based on the coupling coefficients in the coupling matrix, the weighted sum of all individual capability loss values ​​within the operation link is used to obtain the total capability loss prediction value.

[0087] The weighted summation involves multiplying the individual capability loss value for each constraint type within the operational stage by its corresponding weighting coefficient and then summing the results. The weighting coefficients are calculated from the coupling coefficients in the coupling matrix and reflect the overall coupling strength between this constraint type and other constraint types. The total capability loss prediction value is the estimated overall capability loss of the operational stage after comprehensively considering all constraint types and their interactions.

[0088] The system extracts the coupling coefficient corresponding to each constraint type from the coupling matrix. For the i-th constraint type, its weight coefficient is equal to the average of all elements in the corresponding row of the coupling matrix. For example, the physical space constraint is in the first row of the matrix, with elements [1, 1.31, 0.87], and the weight coefficient = (1 + 1.31 + 0.87) / 3 = 1.06. The system obtains the individual capability loss value for this constraint; for example, the individual capability loss value for the physical space constraint is 0.2. The weighted loss value is calculated as 0.2 × 1.06 = 0.212. This calculation is repeated for all constraints involved in the operation, and all weighted loss values ​​are summed to obtain the predicted total capability loss value. For example, if the weighted loss values ​​for three constraints are 0.212, 0.315, and 0.478, the predicted total capability loss value is 0.212 + 0.315 + 0.478 = 1.005.

[0089] S213. Convert the total capability loss prediction value into expected data for real machine tasks, collect the actual data of the target students in the real machine environment, calculate the mean and standard deviation of the total capability loss prediction value of each operation link, and mark the operation links that exceed the mean by a preset multiple of the standard deviation as links to be analyzed.

[0090] Here, "expected data" refers to the student's anticipated performance in a real-world testing environment, calculated based on the predicted total capability loss, such as expected completion time and expected number of errors. "Actual data" refers to the objective data recorded by the student during actual operation in the real-world environment. "Mean" is the arithmetic mean of the predicted total capability loss across all operational stages. "Standard deviation" refers to the dispersion of the predicted total capability loss relative to the mean. "Stages to be analyzed" refers to operational stages with abnormally high predicted total capability loss values, requiring focused attention and in-depth analysis.

[0091] The system calculates the statistical characteristics of the predicted total capacity loss. Assuming the real-machine task includes 5 operational stages, the predicted total capacity loss values ​​are 1.005, 0.756, 2.134, 0.892, and 1.213, respectively. The mean is calculated as (1.005 + 0.756 + 2.134 + 0.892 + 1.213) / 5 = 1.2. The variance is calculated as [(1.005 - 1.2)² + (0.756 - 1.2)² + (2.134 - 1.2)² + (0.892 - 1.2)² + (1.213 - 1.2)²] / 5 = 0.238. The standard deviation is calculated as √0.238 = 0.488. A preset multiplier of 1.5 is set, and the threshold is calculated as mean + 1.5 × standard deviation = 1.2 + 1.5 × 0.488 = 1.932. The system marks operation segments with a total capacity loss prediction value exceeding 1.932 as segments to be analyzed. For example, the prediction value of segment 3, 2.134, exceeds the threshold and is therefore marked as a segment to be analyzed.

[0092] S214. For the analysis stage, calculate the deviation ratio between the expected data and the actual data, and identify the operation stage with the largest number of constraint types and the deviation ratio exceeding the preset threshold as the weak link.

[0093] The deviation ratio refers to the percentage of relative deviation obtained by dividing the difference between the actual data and the expected data by the expected data. The preset threshold is the critical value for judging whether the deviation is significant, for example, set to 30%. The number of constraint types involved refers to how many different types of constraints affect this operation. Weak areas refer to the operation areas where trainees experience the most severe loss of ability when transitioning from a virtual environment to a real-world environment, requiring targeted reinforcement training.

[0094] The system calculates the deviation ratio for each stage to be analyzed. For example, if the expected completion time for stage 3 is 150 seconds and the actual completion time is 210 seconds, the deviation ratio is (210-150) / 150×100%=40%. The system checks if this deviation ratio exceeds a preset threshold of 30%. If it does, it proceeds to the next step. The system counts the number of constraint types involved in this stage. For example, stage 3 involves three types of constraints: physical space constraints, information integrity constraints, and operation feedback delay constraints. The system iterates through all stages to be analyzed, comparing the deviation ratio and the number of constraint types for each stage. For example, stage 5 has a deviation ratio of 35% but only involves two types of constraints, while stage 3 has a deviation ratio of 40% and involves three types of constraints. The system identifies stage 3 as a weak stage and marks it.

[0095] S215. Extract the individual capability loss value corresponding to each constraint type in the weak capability link, sort the individual capability loss values ​​according to their numerical values ​​and then normalize them to obtain the degree of influence of the constraint type.

[0096] The individual capability loss value refers to the capability loss value corresponding to each constraint type in the capability weakness, which has been calculated in S208. Numerical sorting means arranging all individual capability loss values ​​in descending order. Normalization means mapping the sorted loss values ​​to the interval between 0 and 1, so that the influence of different constraint types has a unified measurement standard. Influence degree refers to the relative intensity of the influence of each constraint type on the capability weakness.

[0097] The system extracts all constraint types and their individual capability loss values ​​related to the weak links. For example, link 3 involves physical space constraints (loss value 0.2), information integrity constraints (loss value 0.35), and operational feedback delay constraints (loss value 0.45). The system sorts these loss values ​​from largest to smallest: operational feedback delay constraint 0.45, information integrity constraint 0.35, physical space constraint 0.2. The system calculates the total loss value = 0.45 + 0.35 + 0.2 = 1.0. Each constraint is normalized: operational feedback delay constraint impact = 0.45 / 1.0 = 0.45, information integrity constraint impact = 0.35 / 1.0 = 0.35, physical space constraint impact = 0.2 / 1.0 = 0.2. The system stores the normalized impact values ​​in the constraint impact table.

[0098] S216. Based on the distribution location of weak links and the degree of influence of constraint types, generate the virtual and real ability transfer assessment result value of the target trainees.

[0099] The distribution location refers to the position number of the weak link in the overall operation process, reflecting the stage characteristics of the weak link's occurrence. The degree of influence refers to the relative intensity of the influence of each type of constraint on the weak link, which has been calculated in S215. The virtual-to-real ability transfer assessment result value is a quantitative score that comprehensively evaluates the level of ability retention of trainees when migrating from a virtual environment to a real machine environment. The score ranges from 0 to 100, with higher scores indicating stronger transfer ability.

[0100] The system extracts the location number of the weak link. For example, in 5 operational steps, the weak link is step 3, with a location coefficient of 3 / 5 = 0.6. The system extracts the influence degree of the dominant constraint type. For example, the influence degree of the operational feedback delay constraint is 0.45. The system calculates the base score as 100 × (1 - predicted total capability loss / 5). For example, the predicted total capability loss of step 3 is 2.134, and the base score is 100 × (1 - 2.134 / 5) = 57.32. The system applies location correction: if the weak link is located in the later part of the process (location coefficient > 0.5), the correction coefficient is 1 - 0.1 × location coefficient = 1 - 0.1 × 0.6 = 0.94. The system applies influence degree correction: the correction coefficient is 1 - 0.2 × dominant influence degree = 1 - 0.2 × 0.45 = 0.91. The final evaluation result value = base score × location correction coefficient × influence degree correction coefficient = 57.32 × 0.94 × 0.91 = 49.06, and the system outputs this evaluation result value.

[0101] This step specifically includes:

[0102] The distribution density of weak links in computing power across all operational stages is used to mark continuously distributed weak links as key concern intervals. Based on the degree of influence of each type of constraint, the cumulative impact value of each constraint type within the key concern interval is calculated. Based on the weighted combination of the distribution density and cumulative impact value of the key concern interval, the interval capacity loss coefficient is generated. The interval capacity loss coefficient is then weighted and summed with the total capacity loss prediction value to obtain the virtual-to-real capacity migration assessment result.

[0103] Here, distribution density refers to the concentration of weak links in the operational process, calculated using kernel density estimation to determine the spatial distribution intensity of local weak links. Continuous distribution refers to the distribution of multiple weak links that are adjacent or closely spaced in the operational process, such as weak links appearing consecutively in links 3, 4, and 5. The key focus interval refers to the operational process segment containing continuously distributed weak links, where learner skill transfer problems are most concentrated. The cumulative impact value is the weighted sum of the impact of a specific constraint type on all weak links within the key focus interval; the weights are determined by the relative position of the links within the interval. The interval capability loss coefficient is a quantified value of the overall capability loss within the key focus interval, calculated using a multi-dimensional feature fusion neural network.

[0104] The system constructs a location vector for weak links in all operational stages. Assuming the real-machine task contains 10 operational stages, stages 2, 3, 4, and 7 are marked as weak links, and a binary vector [0, 1, 1, 1, 0, 0, 1, 0, 0, 0] is constructed. The system uses a Gaussian kernel function to calculate the distribution density. For location i, the density value ρ(i) = Σexp(-(ij)² / 2h²), where j represents the locations of all weak links, and h is a bandwidth parameter set to 1.5. For example, the density at position 3 is ρ(3) = exp(-(3-2)² / 2×1.5²) + exp(-(3-3)² / 2×1.5²) + exp(-(3-4)² / 2×1.5²) + exp(-(3-7)² / 2×1.5²) = exp(-0.222) + exp(0) + exp(-0.222) + exp(-3.556) = 0.801 + 1.0 + 0.801 + 0.029 = 2.631. The system calculates the density values ​​at all positions and identifies local peaks. Regions with density values ​​greater than 1.5 times the global average and exceeding 3 consecutive links are marked as key attention intervals. For example, the interval [link 2-link 4] meets the condition. The system extracts the constraint influence degree matrix M of each weak link in the interval, where rows represent links and columns represent constraint types. For example, the influence vector of stage 2 is [0.3, 0, 0.4, 0], that of stage 3 is [0, 0.35, 0.45, 0], and that of stage 4 is [0.25, 0, 0, 0.5]. The system calculates the position weight vector W using the exponential decay function w(k) = exp(-0.5 × |k - k_center|), where k_center is the center position of the interval. For example, if the center of the interval is stage 3, the weight of stage 2 is w(2) = exp(-0.5 × 1) = 0.606, the weight of stage 3 is w(3) = exp(0) = 1.0, and the weight of stage 4 is w(4) = exp(-0.5 × 1) = 0.606. The system calculates the cumulative influence value of each constraint type. For the j-th constraint type, the cumulative influence value is C(j) = Σ(M(i, j) × W(i)), where i traverses all stages within the interval. The cumulative impact value of physical space constraints is C(1) = 0.3 × 0.606 + 0 × 1.0 + 0.25 × 0.606 = 0.182 + 0 + 0.152 = 0.334, the operation feedback delay constraint is C(3) = 0 × 0.606 + 0.45 × 1.0 + 0 × 0.606 = 0.45, the information integrity constraint is C(2) = 0 × 0.606 + 0.35 × 1.0 + 0 × 0.606 = 0.35, and the error tolerance constraint is C(4) = 0 × 0.606 + 0 × 1.0 + 0.5 × 0.606 = 0.303.The system constructs an interval feature vector F=[ρ_max, ρ_mean, interval length, C(1), C(2), C(3), C(4)], for example F=[2.631, 2.1, 3, 0.334, 0.35, 0.45, 0.303]. The system uses a fully connected neural network to calculate the interval capability loss coefficient. The network contains an input layer (7 neurons), hidden layer 1 (16 neurons, ReLU activation), hidden layer 2 (8 neurons, ReLU activation), and an output layer (1 neuron, Sigmoid activation). The system inputs the feature vector into the network and calculates the interval capability loss coefficient λ=0.847 through forward propagation. The system extracts the total capability loss prediction value sequence L=[1.005, 2.134, 1.987, 2.256, 0.756, 0.892, 1.823, 1.213, 0.945, 1.102] for all operation links. The system calculates the weighted entropy H = -Σ(L(i) / Σ L(i)) × log(L(i) / Σ L(i)), where i iterates through all stages, resulting in H = 2.187. The system calculates the evaluation result value for virtual-real skills transfer using the formula: Evaluation result value = 100 × exp(-λ × α - H × β), where α = 0.8 is the interval loss weight coefficient, and β = 0.3 is the overall loss weight coefficient. Substituting the values, the evaluation result value is: 100 × exp(-0.847 × 0.8 - 2.187 × 0.3) = 100 × exp(-0.678 - 0.656) = 100 × exp(-1.334) = 100 × 0.264 = 26.4. The system outputs this evaluation result value as a comprehensive evaluation index for the trainee's virtual-real skills transfer.

[0105] In some embodiments, step S216 is followed by:

[0106] Training scenarios are categorized into basic movement training scenarios, single-skill training scenarios, and comprehensive skill training scenarios based on their operational nature. Basic movement training scenarios are configured with physical space constraints and operational feedback delay constraints. Single-skill training scenarios are configured with information integrity constraints and error tolerance constraints. Comprehensive skill training scenarios are configured with all constraint types, and the parameters of each constraint type are dynamically adjusted based on the target learner's training performance. Based on the virtual-to-real ability transfer assessment results, training scenarios of corresponding difficulty levels are matched to the target learner. A progressive constraint adjustment mechanism is implemented within the training scenarios, including gradually increasing physical space restrictions, reducing operational prompts, shortening operational feedback time, and reducing error tolerance. The adaptation process of the target learner in scenarios of different difficulty levels is recorded, generating personalized training progression paths.

[0107] Among them, basic motion training scenarios refer to training environments focused on single basic operational actions, such as workpiece clamping and tool changing. Single-skill training scenarios refer to training environments targeting specific complete operational procedures, such as a complete tool setting process or program editing process. Comprehensive skill training scenarios refer to training environments for complex tasks involving multiple combinations of operational procedures, such as a complete part machining task. Difficulty level refers to the classification of the complexity of the training scenario, divided into three levels: beginner, intermediate, and advanced, based on the strictness of the constraint parameters. Progressive constraint adjustment mechanism refers to a dynamic adjustment strategy that gradually changes the constraint parameters during training to approximate the real machine environment. Adaptation process refers to the time-series data of changes in the trainee's operational performance as the constraints gradually become stricter. Personalized training progression path refers to a sequence of training stages customized according to the trainee's ability characteristics, progressing from low to high difficulty.

[0108] The system divides the virtual scene library into three sets based on the training objectives. The basic action scene set includes 12 independent action scenes such as fixture operation and tool selection. Each scene has physical space constraints (operation space size, tool placement) and operation feedback latency constraints (action response delay time). For example, the fixture operation scene has an operation space set to 150% of the actual machine size (1200mm × 900mm) and a response latency set to 50% of the actual machine latency (100ms). The single skill scene set includes 8 complete process scenes such as tool setting and program debugging. Each scene has information integrity constraints (number of displayed parameters, level of detail in prompts) and error tolerance constraints (number of allowed errors, number of undo operations). For example, the tool setting scene displays 200% of the actual machine parameters (6 parameters) and allows 3 times the number of errors (3 times). The comprehensive skill scene set includes 5 complex task scenes such as complete machining tasks. Each scene has all four types of constraints set simultaneously. The system matches the scenario difficulty level based on the virtual-to-real capability transfer assessment result value calculated by S216. Assessment result values ​​of 0-30 correspond to beginner difficulty, 31-60 to intermediate difficulty, and 61-100 to advanced difficulty. For example, if a student's assessment result value is 26.4, the system will match them with a beginner difficulty scenario. The system constructs a set of constraint parameter adjustment functions. For physical space constraints, the adjustment function is S(t) = S_true + (S_init - S_true) × exp(-k_s × t), where S_true is the real device space size (800mm × 600mm), S_init is the initial training space size (1200mm × 900mm), k_s is the adjustment rate coefficient (0.05), and t is the cumulative training time in hours. For information integrity constraints, the adjustment function I(n) = I_true + (I_init - I_true) × (1 - n / N)², where I_true is the number of parameters displayed on the real device (3), I_init is the initial number of displayed parameters (6), n is the number of completed training scenarios, and N is the total number of scenarios for this difficulty level (8). For operation feedback latency constraints, the adjustment function D(t) = D_true - (D_true - D_init) × exp(-k_d × t), where D_true is the real device latency (200ms), D_init is the initial latency (100ms), and k_d is the adjustment rate coefficient (0.08). For error tolerance constraints, the adjustment function E(n) = E_init × (1 - n / N), where E_init is the initial allowed number of errors (3), which decreases to 0 when n = N. The system collects operation performance data in real time during student training.For example, after a student completes one hour of training in the first scenario at the beginner difficulty level, the physical space constraint is adjusted to S(1) = 800 + (1200 - 800) × exp(-0.05 × 1) = 800 + 400 × 0.951 = 800 + 380.4 = 1180.4 mm × 885.3 mm. After completing the third scenario, the information integrity constraint is adjusted to I(3) = 3 + (6 - 3) × (1 - 3 / 8)² = 3 + 3 × 0.391 = 3 + 1.173 = 4.173, which means four parameters are displayed. The system records the student's completion time, number of errors, and operational accuracy in each scenario, forming a performance vector P(n) = [T(n), E(n), A(n)]. For example, the performance vector for scenario 1 is [180 seconds, 2 times, 0.85], for scenario 2 it is [165 seconds, 1 time, 0.88], and for scenario 3 it is [158 seconds, 1 time, 0.90]. The system calculates the performance improvement rate R(n) = (P(n) - P(n-1)) / P(n-1). When the average improvement rate of three consecutive scenarios is less than the threshold of 0.05, the system determines that the learner has adapted to the current difficulty level. The system constructs a Markov decision process model to plan the advancement path. The state space S contains four dimensions: the learner's current difficulty level, cumulative training time, performance improvement rate, and current values ​​of constraint parameters. The action space A contains three actions: maintaining the current difficulty level, increasing the difficulty level by one level, and adjusting constraint parameters. The reward function R(s, a) = w1 × performance improvement + w2 × adaptation speed - w3 × number of failures, where w1 = 0.5, w2 = 0.3, and w3 = 0.2. The system uses the Q-learning algorithm to train the strategy network, with a learning rate α=0.1, a discount factor γ=0.9, and an exploration rate ε decaying from 1.0 to 0.1. After 500 simulated training runs, the system generates an advancement path for a learner with an evaluation score of 26.4: Stage 1 involves completing 12 basic action scenarios at beginner difficulty (estimated 15 hours); Stage 2 involves completing the first 4 single-skill scenarios at beginner difficulty (estimated 8 hours); Stage 3 involves advancing to intermediate difficulty and completing the last 4 single-skill scenarios (estimated 10 hours); Stage 4 involves completing the first 2 comprehensive skill scenarios at intermediate difficulty (estimated 12 hours); and Stage 5 involves advancing to advanced difficulty and completing the last 3 comprehensive skill scenarios (estimated 15 hours). The system stores this advancement path and pushes it to the training management module for execution.

[0109] The capability migration prediction system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of a capability migration prediction system in an embodiment of this application.

[0110] It should be noted that, Figure 3 The structure of the capability migration prediction system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0111] like Figure 3 As shown, the capability migration prediction system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 302 or a program loaded from storage portion 308 into Random Access Memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0112] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0113] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0114] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0115] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0116] Specifically, the capability transfer prediction system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the capability transfer prediction method in industrial skills training provided in the above embodiment.

[0117] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the capability migration prediction system described in the above embodiments; or it may exist independently and not incorporated into the capability migration prediction system. The storage medium carries one or more computer programs that, when executed by a processor of the capability migration prediction system, cause the capability migration prediction system to implement the capability migration prediction method in industrial skills training provided in the above embodiments.

[0118] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application 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 of the technical features. 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 this application.

[0119] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for predicting ability transfer in industrial skills training, characterized in that, The method, applied to a capability migration prediction system, includes: Obtain simplified constraint types for the virtual environment relative to the real device. These constraint types include physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints. For each constraint type, a control virtual environment containing the constraint type and an experimental virtual environment removing the constraint type are set up. The difference in the operational performance index of the target student in the control virtual environment and the experimental virtual environment is recorded. The difference in the operational performance index is used as the dependence strength of the target student on the corresponding constraint type. Extract the constraint types involved in each operation step in the real machine task, determine the individual capability loss value based on the dependency strength corresponding to each constraint type and the difference multiple between the real machine and the virtual environment, calculate the coupling coefficient between different constraint types in the operation step, and obtain the total capability loss prediction value of the operation step by weighted summing of all individual capability loss values ​​and their corresponding coupling coefficients. The predicted total capability loss is converted into expected data for the real-device task. The actual data of the target student in the real-device environment is collected. Based on the deviation data between the expected data and the actual data, the virtual-to-real capability transfer assessment result value of the target student is generated.

2. The method according to claim 1, characterized in that, The step of extracting the constraint types involved in each operation step of the real machine task, and determining the individual capability loss value based on the dependency strength corresponding to each constraint type and the difference multiple between the real machine and the virtual environment, specifically includes: Obtain the operation flow of the real device task and divide the operation flow into multiple operation steps; For each operation step, analyze one or more of the constraint types involved in the operation step, including physical space constraints, information integrity constraints, operation feedback delay constraints, and error tolerance constraints. Obtain the reference baseline value for each constraint type in the real machine environment and the corresponding actual value in the virtual environment; Calculate the difference factor for each constraint type based on the ratio of the reference value to the actual value; The preliminary loss value is calculated based on the dependence strength and the difference factor. The initial loss value is normalized based on the standard completion time of the operation step to obtain the individual capability loss value.

3. The method according to claim 1, characterized in that, The step of calculating the coupling coefficients between different constraint types within the operational phase, and then weighting and summing all individual capacity loss values ​​with their corresponding coupling coefficients to obtain the predicted total capacity loss value for the operational phase, specifically includes: For each pair of constraint types in the operation, record the first operational performance index of the target student in the experimental virtual environment where two constraint types are removed simultaneously. The first operational performance index is compared with the second operational performance index when a single constraint type is removed, and the coupling coefficient between the two constraint types is calculated. Construct a coupling matrix that includes all constraint types within the aforementioned operational phase; Based on the coupling coefficients in the coupling matrix, the weighted sum of all individual capability loss values ​​within the operational step is used to obtain the predicted total capability loss value.

4. The method according to claim 1, characterized in that, The step of generating the virtual-to-real ability transfer assessment result value of the target trainee based on the deviation data between the expected data and the actual data specifically includes: Calculate the mean and standard deviation of the total capacity loss prediction for each operation step, and mark the operation steps that exceed the mean by a preset multiple of the standard deviation as the steps to be analyzed. For the process to be analyzed, the deviation ratio between the expected data and the actual data is calculated, and the process with the largest number of constraint types and a deviation ratio exceeding a preset threshold is identified as the weakest link. Extract the individual capability loss value corresponding to each constraint type in the weak capability link, sort the individual capability loss values ​​according to their numerical values ​​and then normalize them to obtain the degree of influence of the constraint type. Based on the distribution of the weak links and the degree of influence of the constraint types, the virtual and real ability transfer assessment result value of the target trainee is generated.

5. The method according to claim 4, characterized in that, The step of generating the virtual-to-real skills transfer assessment result value of the target trainee based on the distribution location of the weak links and the degree of influence of the constraint type specifically includes: Calculate the distribution density of the weak links in all operational processes, and mark continuously distributed weak links as key areas of concern; Based on the degree of influence of the aforementioned constraint types, calculate the cumulative impact value of each constraint type within the key focus interval; Based on the weighted combination of the distribution density of the key focus intervals and the cumulative impact value, an interval capacity loss coefficient is generated; The virtual-real capability migration assessment result is obtained by weighting and summing the interval capability loss coefficient and the total capability loss prediction value.

6. The method according to claim 1, characterized in that, After the step of generating the virtual-to-real ability transfer assessment result value of the target trainee based on the deviation data between the expected data and the actual data, the method further includes: Build a virtual training scenario library and set up training scenarios with multiple difficulty levels for each type of constraint. Based on the virtual-real ability transfer assessment results, training scenarios of corresponding difficulty levels are matched for the target learners; A progressive constraint adjustment mechanism is set in the training scenario. The constraint adjustment mechanism includes gradually increasing physical space restrictions, reducing operation prompt information, shortening operation feedback time, and reducing error tolerance. Record the adaptation process of the target learners in scenarios with different difficulty levels, and generate personalized training advancement paths.

7. The method according to claim 6, characterized in that, The step of constructing a virtual training scenario library, which involves setting multiple difficulty levels for training scenarios for each constraint type, specifically includes: The training scenarios are categorized according to their operational nature into basic movement training scenarios, single skill training scenarios, and comprehensive skill training scenarios. In the basic motion training scenario, physical space constraints and operation feedback delay constraints are set; In the single-skill training scenario, information integrity constraints and error tolerance constraints are set; In the comprehensive skills training scenario, all constraint types are set, and the parameters of each constraint type are dynamically adjusted according to the training performance of the target trainee.

8. A capability transfer prediction system, characterized in that, The capability migration prediction system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the capability migration prediction system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the capability migration prediction system, the capability migration prediction system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the capability migration prediction system, the capability migration prediction system performs the method as described in any one of claims 1-7.

Images