Intelligent cloud control method and system supporting multi-species practical training

Abstract

This invention relates to the field of smart cloud training technology, and discloses a smart cloud control method and system supporting multi-skill training. The method includes: acquiring global data of multi-skill collaborative training sessions; calculating the estimated sequence of equivalent contact resistance values ​​for each shared electrical node based on the empirical formula for contact degradation; injecting the estimated sequence into a digital twin simulation entity to generate a degradation impact propagation vector; performing dual-path causal tracing for error events of trainees in each downstream skill category, calculating the correlation between human operation propagation and hardware degradation propagation respectively; classifying error events based on competitive comparison results to generate an error propagation label matrix; updating the independent capability defect profile vector of each skill category trainee; and generating personalized training paths based on the vector and synchronizing them to the smart cloud training platform.

Description

Technical Field

[0001] This invention relates to the field of smart cloud training technology, and more specifically, to a smart cloud control method and system that supports multi-task training. Background Technology

[0002] In multi-skill collaborative training scenarios, trainees from different skill classes simultaneously complete interconnected training sub-tasks online. Input-output dependencies exist between these sub-tasks, and multiple trainees share certain electrical nodes. Existing intelligent cloud control methods use a multi-skill task dependency graph to trace whether operational errors from upstream trainees propagate to the operating environments of downstream trainees, thereby identifying passive errors and eliminating their interference with trainee competency assessment. Simultaneously, the contact status of shared electrical nodes at each training station is independently monitored, triggering threshold alarms for abnormal contact resistance.

[0003] However, when shared electrical nodes by multiple trainees undergo gradual degradation—for example, the contact resistance of common bus terminals gradually increases due to repeated operation by different trainees—the signal attenuation or increased noise caused by degradation can lead to abnormal readings in the testing and calibration tasks of downstream trainees. Existing intelligent cloud control methods have the following shortcomings: First, error propagation detection only traces human operational errors as the propagation source, failing to identify non-operational error sources such as gradual changes in hardware contact status. Second, contact status monitoring only outputs terminal health alarms, failing to establish a causal relationship between the degradation state and errors by downstream trainees. These two disconnects result in downstream trainees' passive errors caused by hardware degradation being incorrectly attributed to their own skill deficiencies or operational errors by upstream trainees. Consequently, the cloud platform pushes unnecessary reinforcement training paths, reducing the overall control accuracy and efficiency of multi-trainee collaborative training. Summary of the Invention

[0004] This invention provides a smart cloud control method and system that supports multi-skill training, solving the technical problems in related technologies such as inaccurate attribution of error events in multi-skill collaborative training scenarios, interference of hardware degradation with trainee ability assessment, and ineffective collaborative scheduling of personalized training paths.

[0005] This invention discloses a smart cloud control method that supports multi-task training, including: acquiring global data of multi-task collaborative training sessions, and calculating the equivalent contact resistance estimate sequence of each shared electrical node based on the contact degradation empirical formula; The estimated sequence of equivalent contact resistance is injected into the cloud-based digital twin simulation entity to perform signal transmission simulation and generate a degradation effect propagation vector. A dual-path causal tracing was performed on the error events of trainees in each downstream job category to obtain the correlation between human operation propagation and hardware degradation propagation. Based on the competitive comparison between the correlation between human operation propagation and the correlation between hardware degradation propagation, the error event is attributed to autonomous error, human propagation error or degradation propagation error, and an error propagation label matrix is ​​generated. Based on the error propagation marker matrix, human-caused errors and deterioration propagation errors are eliminated, generating independent capability defect profile vectors for trainees in each job category, and generating personalized training paths based on these vectors and synchronizing them to the smart cloud training platform.

[0006] Furthermore, the global data includes the timing data of the operation behavior of trainees in each trade, the multi-trade task dependency graph, the operation history of each shared electrical node, and the feedback value sequence of the tightening torque sensor; The empirical formula for contact degradation expresses the estimated equivalent contact resistance at the current sampling time as: the initial contact resistance value of the shared node multiplied by a degradation factor, wherein the degradation factor is calculated as follows:

[0007] in, The degradation factor is... The degradation rate coefficient is... This represents the cumulative number of operations up to the current moment. The number of operations is the index. The torque attenuation factor is... This is the sliding average of the tightening torque feedback value up to the current moment. It is a natural constant; The degradation rate coefficient, the number of operations index, and the torque attenuation factor are empirical parameters that are pre-calibrated through experimental fitting based on the terminal material type and structural specifications. They are stored in the smart cloud training platform in the form of a parameter table according to the terminal material type and structural specifications, and are retrieved from the table when calculating the estimated sequence of equivalent contact resistance values ​​for each shared node.

[0008] Furthermore, the step of injecting the equivalent contact resistance estimate sequence into the cloud-based digital twin simulation entity also includes: Perform thermal accumulation simulation, calculate the instantaneous heating power of each shared node based on the Joule heating formula, and perform time step iteration by combining material heat capacity parameters and heat dissipation coefficient to obtain the cumulative temperature rise curve of each shared node; The signal transmission simulation includes: based on the circuit topology of the branch where each shared node is located in the digital twin simulation entity, substituting the estimated value of the equivalent contact resistance at the current moment into the voltage division or current division equivalent model of the branch to obtain the signal attenuation at the downstream measurement point; based on the positive correlation between the thermal noise power of the contact resistance and the estimated value of the equivalent contact resistance, calculating the increment of the root mean square value of the equivalent noise voltage superimposed on the downstream signal as the noise increment. Each component of the degradation effect propagation vector corresponds to the degree of signal quality degradation of the downstream measurement channel by each shared node at each time.

[0009] Furthermore, the first path in the dual-path causal attribution is a human-operated propagation attribution, including: Tracing back along the reverse edges of the multi-task dependency graph, we find the operation outputs of upstream trainees who have a task dependency relationship with the current error event; The temporal correlation is calculated using a natural exponential function with the negative of the product of the time decay coefficient and the time interval between the upstream abnormal output time and the current error event occurrence time as the exponent. The semantic matching degree is calculated based on the set of parameter types of the upstream abnormal output and the set of input parameter types involved in the current error event, and is obtained using the Jaccard similarity coefficient algorithm. The temporal relevance and the semantic matching degree are weighted and fused to obtain the human operation propagation relevance, wherein both weight coefficients are non-negative real numbers and their sum is one.

[0010] Furthermore, the second path in the dual-path causal attribution is hardware degradation propagation attribution, including: Based on the degradation effect propagation vector, the downstream signal attenuation and noise increment of the shared node corresponding to the time of the error event are extracted; The direction and amplitude of the deviation of the downstream measurement point signal output by the digital twin simulation entity at this moment are compared with the direction and amplitude of the deviation of the abnormal reading actually observed by the trainee in the current error event. When the deviation directions are consistent and the amplitude ratio falls within the preset matching range, the hardware degradation propagation correlation is assigned a value of 1; otherwise, it is assigned a value of 0. The preset matching interval is a symmetrical interval centered at 1. Its upper and lower limits are determined by offline simulation calibration based on the signal modeling error range of the digital twin simulation entity for the node, and are stored in the smart cloud training platform in the form of interval endpoint pairs.

[0011] Furthermore, when the correlation of hardware degradation propagation is higher than the correlation of human operation propagation and exceeds a preset causal threshold, the error event is attributed to a degradation propagation error, and the source shared node causing the degradation is marked; the preset causal threshold is a preset real number greater than 0 and less than 1, which is preset and stored in the smart cloud training platform according to the balance requirements of attribution misclassification rate and missed classification rate in the training scenario; it also includes: The torque deviation of each type of trainee in the most recent operation is normalized to the mean based on the range. Calculate the responsibility contribution ratio of each job type trainee operating the source shared node. The responsibility contribution ratio is the sum of the trainee's historical operation frequency ratio and the normalized torque deviation multiplied by the torque deviation weight coefficient, divided by the sum of the corresponding values ​​of all trainees who have operated the node. The torque deviation weighting coefficient is pre-calibrated experimentally based on the terminal material type.

[0012] Furthermore, the generation of independent competency deficiency profile vectors for trainees in each job category includes: The autonomous errors of trainees in each type of work are statistically analyzed according to the knowledge point dimension and the operational skill dimension involved. Each component of the independent ability defect profile vector corresponds to the degree of defect in each knowledge point dimension and operational skill dimension. The contribution of each trainee to the degradation of the shared node is included in the operational standardization scoring dimension. The operational contribution is obtained by accumulating the responsibility contribution ratio corresponding to the degradation propagation error in the error propagation label matrix.

[0013] Furthermore, the generation of personalized training paths also includes: Based on the independent capability defect profile vector, candidate training task nodes corresponding to the defect dimension of each type of trainee are matched in the practical training task knowledge graph to generate a set of candidate training paths. Extract the shared training resource occupancy time window and cross-job task prerequisites involved in each candidate training path; Using a multi-agent coordination and scheduling algorithm based on constraint satisfaction, under the constraints of no temporal conflict in shared resources and the order of task dependencies between job types, the candidate training paths of all job types trainees are jointly solved, and the globally optimal path combination scheme that maximizes the rate of reduction of the average independent capability defect of the group as a whole within the current collaborative training session period is output to replace the set of candidate training paths. The multi-agent coordination scheduling algorithm based on constraint satisfaction takes the shared training resource occupation time window and cross-job task prerequisite dependency conditions of each candidate training path as input, takes the non-overlapping time of shared resource occupation and the satisfaction of the completion order of prerequisite tasks between jobs as hard constraints, and takes maximizing the reduction rate of the average independent ability defect profile vector of all trainees in the current collaborative training session as the optimization objective, and outputs the global optimal path combination scheme.

[0014] Furthermore, it also includes: Perform trend extrapolation analysis on the cumulative temperature rise curves of each shared node, perform linear or polynomial extrapolation based on the assumption that the current multi-task operation mode continues, predict the estimated remaining time for each node to reach the material critical temperature, generate a thermal degradation countdown vector, and synchronize the thermal degradation countdown vector to the smart cloud training platform to trigger maintenance prompts for the shared nodes. After any trainee completes a subtask, the error propagation marker matrix, the shared resource occupancy status, and the estimated equivalent contact resistance of each node are updated in real time. This triggers dynamic re-coordination of the paths of the affected trainees and records the health decay trajectory and deterioration responsibility distribution of each shared node to the cloud-based equipment maintenance database.

[0015] This invention provides a smart cloud control system for multi-task collaborative training, comprising: The global data acquisition and contact resistance calculation module is used to acquire global data from multi-trade collaborative training sessions and calculate the estimated sequence of equivalent contact resistance values ​​for each shared electrical node based on the empirical formula for contact deterioration. The digital twin simulation module is used to inject the estimated sequence of equivalent contact resistance values ​​into the cloud-based digital twin simulation entity, perform signal transmission simulation, and generate a degradation effect propagation vector. The dual-path causal tracing module is used to perform dual-path causal tracing on the error events of trainees in each downstream job category, and obtain the correlation between human operation propagation and hardware degradation propagation respectively. The attribution classification module is used to attribute the error event to autonomous error, human error propagation error, or hardware degradation propagation error based on a competitive comparison between the human operation propagation correlation and the hardware degradation propagation correlation, and to generate an error propagation label matrix. The competency defect profile generation module is used to eliminate human-induced errors and deterioration-induced errors based on the error propagation marker matrix, and generate independent competency defect profile vectors for trainees in each job category. The training path generation and synchronization module is used to generate personalized training paths based on the independent capability defect profile vector and synchronize the control results to the smart cloud training platform.

[0016] The present invention has the following beneficial effects: Firstly, by using an empirical formula for contact degradation, the operational history of shared electrical nodes and the feedback values ​​from tightening torque sensors are transformed into a sequence of estimated equivalent contact resistance values. This sequence is then injected into a digital twin simulation entity to generate a degradation impact propagation vector, thereby incorporating the gradual degradation of hardware contact conditions into a causal tracing framework for error propagation. Compared to existing methods that only trace human operational errors, this invention can independently identify the non-operational error source of gradual hardware degradation.

[0017] Secondly, by employing a dual-path causal attribution approach and competitive comparison, the correlation between human error propagation and hardware degradation propagation is simultaneously assessed for each downstream trainee's error events. Error events are then categorized into three types: autonomous errors, human-propagated errors, and degradation-propagated errors. Passive errors caused by hardware degradation are no longer attributed to trainees' own skill deficiencies or operational errors by upstream trainees, thus preventing the cloud platform from pushing unnecessary reinforcement training paths to affected trainees.

[0018] Third, by using the decoupled independent capability defect profile vector as the input for path generation, the personalized training paths of trainees in each trade reflect only their true autonomous capability defects. This enables conflict-free allocation of shared resources and satisfaction of task dependency order among trades in the joint constraint satisfaction scheduling, thereby simultaneously improving the attribution accuracy and global path coordination efficiency of the intelligent cloud platform's control over multi-trade collaborative training. Attached Figure Description

[0019] Figure 1 This is a flowchart of a smart cloud control method supporting multi-task training provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the trend of the estimated equivalent contact resistance of the BT-02 node as a function of sampling time, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram illustrating the coordinated change between the cumulative number of operations and the moving average torque of the BT-02 node provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the changes in signal attenuation and noise increment of the IN-07 downstream channel over time, as provided in an embodiment of the present invention. Figure 5 This is a schematic diagram of the cumulative temperature rise curve of the BT-02 node digital twin thermal simulation provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the correlation score comparison of the dual-path causal tracing of the ERR-IN07-04 error event provided in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the contribution ratio of EL-03 and IN-07 to the degradation of the BT-02 node, provided in an embodiment of the present invention. Figure 8 This is a schematic diagram of the comparison of student independent ability deficiency profile vectors provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the scatter distribution of the estimated equivalent contact resistance and downstream signal attenuation provided in the embodiments of the present invention. Detailed Implementation

[0020] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0021] Example A smart cloud control method supporting multi-task training according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain global data for multi-task collaborative training sessions and calculate the sequence of estimated equivalent contact resistance values ​​for shared nodes.

[0022] Obtain global data for the current multi-skill collaborative training session in the smart cloud training platform. Global data includes the timing data of each trainee's operational behavior, a multi-skill task dependency graph, historical operation records of each shared electrical node, and a sequence of tightening torque sensor feedback values.

[0023] Based on the empirical formula for contact degradation, a sequence of estimated equivalent contact resistance values ​​for each shared node is calculated. The empirical formula for contact degradation is expressed as:

[0024] in, For a moment The estimated equivalent contact resistance; This is the current sampling time; This is the initial contact resistance value of the shared node; Deadline The cumulative number of operations is obtained from the insertion / removal or tightening operation count in the operation history. Deadline The sliding average value of the tightening torque sensor feedback value; This is the degradation rate coefficient; This is an index representing the number of operations. This is the torque attenuation factor; This is a natural constant. The above... , , These are empirical parameters pre-calibrated based on the terminal material type and structural specifications.

[0025] It should be noted that in the above empirical formula for contact degradation, The term characterizes the cumulative damage effect of repeated operations on the microstructure of the contact surface; the more operations, the greater the increase in contact resistance. The term characterizes the compensating effect of the tightening torque on the tightness of the contact surface; the greater the torque, the tighter the contact surface fits, and the stronger the suppression of degradation increment. The product relationship between the two terms allows the empirical formula for contact degradation to simultaneously reflect the coupling effect between cumulative damage and tightening compensation.

[0026] Calculate the time step for each shared node according to the sampling time step. Generate a sequence of estimated equivalent contact resistance values ​​for each shared node.

[0027] Step 2: Perform digital twin simulation based on the estimated sequence of equivalent contact resistance to generate a degradation effect propagation vector.

[0028] The estimated equivalent contact resistance of each shared node is injected into the corresponding node impedance parameters of the cloud-based digital twin simulation entity.

[0029] Perform thermal accumulation simulation. Calculate the instantaneous heat generation power of each node based on the Joule heating formula, and perform time-step iterations by combining material heat capacity parameters and heat dissipation coefficients to obtain the cumulative temperature rise curves of each shared node.

[0030] Perform signal transmission simulation. Calculate the downstream signal attenuation and noise increment caused by changes in contact resistance at each shared node.

[0031] It should be noted that in the above signal transmission simulation, the downstream signal attenuation is calculated as follows: based on the circuit topology of the branches where each shared node is located in the digital twin simulation entity, the current signal attenuation is calculated... Substituting the equivalent voltage or current division model of this branch, the signal amplitude deviation at the downstream measurement point of this node is obtained. The noise increment is calculated as follows: based on the thermal noise power of the contact resistance and... The positive correlation is used to calculate the increment of the root mean square value of the equivalent noise voltage superimposed on the downstream signal under the current contact state.

[0032] The downstream signal attenuation and noise increment of each shared node at each time step are summarized to generate a degradation effect propagation vector. Each component of the degradation effect propagation vector corresponds to the degree of signal quality degradation of the downstream measurement channel by each shared node at each time step.

[0033] Step 3: Perform dual-path causal tracing for error events of trainees in each downstream job category.

[0034] Obtain the subsequences of operations marked as errors from the time-series data of each downstream trainee's operational behavior. Perform two-path causal tracing for each error event of each downstream trainee.

[0035] The first path involves tracing the propagation of human actions. Following the reverse edges of the multi-task dependency graph, the operation outputs of upstream trainees that have a task dependency relationship with the current error event are traced. The temporal relevance and semantic matching degree between the upstream operation outputs and the current error event are calculated. Temporal relevance... Based on the time interval between the upstream abnormal output time and the current error event occurrence time. The calculation is obtained using an exponential decay function:

[0036] in, For time-series correlation; It is a natural constant; The time decay coefficient is pre-calibrated based on the typical response time of task transfer between different job types. The time interval between the upstream abnormal output time and the current error event occurrence time. The smaller The larger the value, the higher the temporal correlation.

[0037] Semantic matching degree Parameter type set based on upstream anomaly output The set of input parameter types involved in the current error event The similarity coefficient was obtained using the Jaccard similarity coefficient algorithm. The range of values ​​is The larger the value, the greater the overlap in parameter types between the two sets.

[0038] By weighting and fusing temporal relevance and semantic matching, the relevance of human-operated propagation is obtained. :

[0039] in, The relevance of information disseminated through human manipulation; and For the preset weighting coefficients, satisfy ; For time-series correlation; This represents the semantic matching degree.

[0040] The second approach involves tracing the propagation of hardware degradation. Based on the degradation impact propagation vector, the downstream signal attenuation and noise increment at the shared node corresponding to the time of the error event are extracted. The degree of signal degradation reflected by the downstream signal attenuation and noise increment is matched with the deviation amplitude of the abnormal readings in the current error event to obtain the correlation of hardware degradation propagation. .

[0041] It should be noted that the above matching calculation refers to comparing the direction and amplitude of the deviation signal from the downstream measurement point output by the digital twin simulation entity at that moment with the direction and amplitude of the deviation of the abnormal reading actually observed by the student in the current error event. When the deviation directions are consistent and the amplitude ratio falls within the preset matching interval, Take 1; otherwise, Set to 0. The above-mentioned preset matching interval is pre-calibrated based on the signal modeling error range of the digital twin simulation entity for this node, representing the closed interval formed by the acceptable lower and upper limits of the amplitude ratio.

[0042] Step 4: Attribution classification of error events based on dual-path relevance to generate an error propagation label matrix.

[0043] The correlation between human intervention and propagation of each error event Correlation with hardware degradation propagation Conduct a competitive comparison.

[0044] when Higher than Furthermore, if the error exceeds a preset causal threshold, the error event is attributed to a degradation propagation error, and the source shared node that triggered the degradation is identified. Simultaneously, based on the percentage of operations performed by each trainee on the source shared node in the operation history and the torque deviation of the most recent operation, the responsibility contribution ratio of each trainee on the source shared node is calculated. Before calculating the responsibility contribution ratio, the torque deviation of each trainee's most recent operation is normalized to the mean based on the range, scaling it to... The interval eliminates the influence of torque dimensions on the weighted summation operation. Specifically, for trainees in various trades... Proportion of responsibility contribution to source-shared nodes The calculation is as follows:

[0045] in, For trainees of various trades The proportion of responsibility contribution to the source-sharing nodes; Index of target students for calculating the current proportion of responsibility contribution; For students The historical number of operations performed on this node; This is the sum of the historical number of operations performed by all students who operated this node; For students The dimensionless value of the absolute value of the torque deviation of the most recent operation after normalization; The torque deviation weighting coefficient is pre-calibrated based on the terminal material type; This represents the summation over all students who have interacted with that node. This represents the total number of students who have operated this node. For summation index.

[0046] when Higher than Furthermore, if the error exceeds a preset causal threshold, the error event will be attributed to human error propagation.

[0047] When the correlation between the two paths does not exceed the preset causal threshold, the error event is attributed to an autonomous error.

[0048] The above-mentioned preset causal threshold value range is: Based on the correlation distribution of various causal paths in historical data during the system debugging phase, it is pre-calibrated to distinguish between strong causal associations and weakly correlated noise.

[0049] Summarize the attribution results of all error events to generate an error propagation label matrix. The rows of the error propagation label matrix correspond to trainees of each job type, and the columns correspond to each error event. The elements of the error propagation label matrix are labeled as one of three categories: autonomous error, human-propagated error, or degradation-propagated error.

[0050] Step 5: Update the independent competency defect profile vectors of trainees in each job type based on the error propagation label matrix.

[0051] Based on the error propagation labeling matrix, deterioration propagation errors and human-propagated errors are removed from the corresponding trainees' error sets, retaining only autonomous errors. Autonomous errors for each job type are statistically analyzed according to the knowledge point dimension and operational skill dimension involved, generating an independent competency deficiency profile vector for each job type trainee. Each component of the independent competency deficiency profile vector corresponds to the degree of deficiency in each knowledge point dimension and operational skill dimension.

[0052] Simultaneously, each trainee's contribution to the degradation of shared nodes will be included in the operational standardization scoring dimension. The operational contribution is based on the proportion of responsibility for degradation propagation errors in the error propagation labeling matrix. Cumulative earnings.

[0053] Step 6: Generate personalized training paths based on independent capability defect profile vectors and synchronize the control results to the smart cloud training platform.

[0054] Based on the independent capability defect profile vector, candidate training task nodes corresponding to the defect dimensions of trainees in each type of work are matched in the knowledge graph of practical training tasks, and a set of candidate training paths is generated.

[0055] The personalized training paths of trainees in each trade, the proportion of three types of error attribution, and the shared node degradation warning information are synchronized to each workstation terminal of the smart cloud training platform.

[0056] In this embodiment of the application, in order to improve the coordination between paths when trainees from different trades share practical training resources, the following steps are also included in step 6: Extract the shared training resource occupancy time window and cross-job task prerequisites for each candidate training path. Utilize a constraint-based multi-agent coordination scheduling algorithm to jointly solve for the candidate training paths for all trainees across job types. This algorithm aims to maximize the rate of decrease in the average independent capability defect of the group as a whole within the current collaborative training session cycle. Hard constraints include the absence of temporal conflicts in shared resources and the fulfillment of task dependencies between job types. The inputs are the resource occupancy time window and task prerequisites for each candidate training path, and the output is the globally optimal path combination scheme. Replace the original candidate training path set with the globally optimal path combination scheme and synchronize it to each workstation terminal.

[0057] In this embodiment of the application, in order to predict the thermal safety risks of shared nodes in advance during the training process, the following steps are also included in step 2 based on obtaining the cumulative temperature rise curve: Perform trend extrapolation analysis on the cumulative temperature rise curves of each shared node. Based on the assumption of the current multi-tasking operation mode, perform linear or polynomial extrapolation on the cumulative temperature rise curves to predict the estimated remaining time for each node to reach the material's critical temperature, generating a thermal degradation countdown vector. Synchronize the thermal degradation countdown vector to the smart cloud training platform to trigger maintenance prompts for the shared nodes.

[0058] In this embodiment of the application, in order to dynamically update the control results as the training process progresses, the following steps are further included after synchronization is completed in step 6: After any trainee completes a subtask, the error propagation marker matrix, shared resource occupancy status, and estimated equivalent contact resistance of each node are updated in real time. Based on the updated error propagation marker matrix, shared resource occupancy status, and estimated equivalent contact resistance, the paths of affected trainees are dynamically re-coordinated. Simultaneously, the health degradation trajectory and deterioration responsibility distribution of each shared node are recorded in the cloud-based equipment maintenance database.

[0059] This implementation transforms the operational history of shared electrical nodes and the feedback values ​​from tightening torque sensors into a sequence of estimated equivalent contact resistance values ​​using an empirical formula for contact degradation. This sequence of estimated equivalent contact resistance values ​​is then injected into a digital twin simulation entity to generate a degradation impact propagation vector. This incorporates the gradual degradation of hardware contact states into the causal tracing framework of error propagation, enabling the independent identification of hardware degradation as a non-operational error source. This overcomes the limitation of existing methods that only trace human operational errors as the source of propagation.

[0060] By employing a dual-path causal attribution approach and competitive comparison, the correlation between human error propagation and hardware degradation propagation is simultaneously assessed for each downstream trainee's error events. Based on the competitive comparison results, error events are categorized into three types: autonomous errors, human error propagation errors, and degradation propagation errors. This prevents passive errors caused by hardware degradation from being attributed to trainees' own skill deficiencies or operational errors by upstream trainees, thereby avoiding the cloud platform from pushing unnecessary reinforcement training paths to affected trainees.

[0061] By using the decoupled independent capability defect profile vector as input for path generation, the personalized training paths of trainees in each trade reflect only their true autonomous capability defects. This enables conflict-free allocation of shared resources and satisfaction of task dependency order among trades in the joint constraint satisfaction scheduling, thereby simultaneously improving the attribution accuracy and global path coordination efficiency of the intelligent cloud platform's control over multi-trade collaborative training.

[0062] The following is an example of an application of the present invention, such as... Figure 2-9 As shown, the implementation process is as follows: A vocational skills training institution conducted collaborative training on the debugging of automated control systems on a smart cloud training platform. In B month of 20XX, an electrician (EL-03) and an instrumentation technician (IN-07) were online simultaneously, collaborating to complete the wiring and calibration of a PLC control circuit. The two trainees shared a common bus terminal node (BT-02). The electrician was responsible for upstream wiring and power configuration, while the instrumentation technician was responsible for downstream sensor signal calibration. The BT-02 terminal had accumulated some contact degradation due to repeated plugging and unplugging operations by previous trainees. The training platform's cloud server cluster hosted digital twin simulation entities, and tightening torque sensors were deployed at each workstation to collect real-time feedback values ​​of the trainees' operating torque.

[0063] Step 1: Obtain global data and calculate the sequence of estimated equivalent contact resistance values ​​for shared nodes; The platform collects the operation history and torque sensor feedback value sequence of the BT-02 node. As of the 4th sampling time in this training session, the cumulative number of operations and the moving average torque of BT-02 are shown in Table 1. The initial contact resistance of the BT-02 node is also shown. The empirical parameters calibrated by the terminal material are as follows: , , .

[0064] At each moment and Substituting into the empirical formula for contact degradation

[0065] The estimated equivalent contact resistance was calculated at each time step, and the results are listed in Table 1.

[0066] Table 1. Estimated series of equivalent contact resistance values ​​for BT-02 node.

[0067] As the cumulative number of operations increases, the moving average torque decreases. The cumulative damage effect is enhanced. With the weakening of the fastening compensation, the estimated equivalent contact resistance of BT-02 gradually increased from 0.11Ω at time 1 to 0.15Ω at time 4, showing a gradual deterioration trend.

[0068] Step 2: Perform digital twin simulation based on the estimated sequence of equivalent contact resistance to generate a degradation effect propagation vector; The time intervals in Table 1 Inject the node impedance parameters corresponding to BT-02 into the digital twin simulation entity.

[0069] Thermal accumulation simulation was performed: Based on the Joule heating formula, and combined with the material heat capacity parameters and heat dissipation coefficient of the BT-02 terminal, a time-step iteration was performed to obtain the cumulative temperature rise curve of BT-02. At time 4, the cumulative temperature rise reached 11.3K, which was higher than 6.8K at time 1, and the temperature rise curve showed an accelerating upward trend.

[0070] Perform trend extrapolation analysis on the cumulative temperature rise curve. Based on the assumption that the current operating mode continues, perform polynomial extrapolation to predict that the estimated remaining time for BT-02 to reach the material critical temperature (85°C, current base temperature 23°C) is about 47 minutes. Generate a thermal degradation countdown vector and synchronize it to the platform maintenance prompt module.

[0071] Perform signal transmission simulation: Simulate the signal transmission at each time step. Substituting the voltage divider equivalent model of the branch where BT-02 is located, the signal amplitude deviation at the downstream measurement point of IN-07 is calculated; based on the contact resistance thermal noise power and The positive correlation between the two signals was used to calculate the root mean square increment of the equivalent noise voltage superimposed on the downstream signal. The simulation results are shown in Table 2.

[0072] Table 2. Propagation vector of BT-02 node degradation impact (IN-07 downstream measurement channel)

[0073] Each component of the degradation effect propagation vector varies with The signal attenuation increased monotonically, reaching 5.70mV at time 4, and the noise increment reached 0.31mV, providing quantitative evidence for subsequent causal tracing.

[0074] Step 3: Perform dual-path causal tracing for error events of trainees in each downstream job category; IN-07 An error event (ERR-IN07-04) occurred near time 4: When zero-point calibration was performed on a current signal from 4 to 20mA, the reading was abnormally low, with a deviation of approximately 5.30mV and a negative attenuation direction.

[0075] First Path – Human Operation Propagation Tracing: Tracing back along the reverse edge of the multi-task dependency graph, EL-03 completes upstream power supply configuration output at time 3, and the time interval between this and ERR-IN07-04 is... Minutes. Time decay coefficient. (calibrated according to the typical response time between different types of work).

[0076]

[0077] The parameter type set of the upstream anomaly output of EL-03 is as follows: Voltage amplitude, phase sequence The set of input parameter types involved in ERR-IN07-04 is as follows: Current range, zero bias If the intersection of the two sets is empty, the Jaccard similarity coefficient algorithm is used to obtain... .

[0078] Take preset weights , Weighted fusion yields:

[0079] The second path—tracing the propagation of hardware degradation: Extracting the downstream signal attenuation of 5.70mV (negative) and noise increment of 0.31mV corresponding to time 4 of BT-02. The deviation direction of the digital twin simulation entity's output is negative attenuation, consistent with the negative deviation direction of the abnormal reading observed by the student ERR-IN07-04; the amplitude ratio of the simulation output amplitude of 5.70mV to the actual deviation amplitude of 5.30mV is 1.08, falling within the preset matching range. Within, both direction and amplitude satisfy the matching condition, therefore .

[0080] The results of dual-path tracing are summarized in Table 3.

[0081] Table 3 Results of Dual-Path Causal Origin Tracing of ERR-IN07-04 Error Event

[0082] Step 4: Attribution and classification of error events based on dual-path relevance, generating an error propagation label matrix; The preset causal threshold is 0.60. ,and Exceeding the causality threshold of 0.60, ERR-IN07-04 was attributed to a deterioration propagation error, and the source shared node was labeled as BT-02.

[0083] Calculate the contribution ratio of EL-03 and IN-07 to BT-02. EL-03 historical operation count. IN-07 historical operation count , The absolute value of the torque deviation during the most recent operation of EL-03 was 0.42 N·m, while that of IN-07 was 0.28 N·m. After range normalization (range of 0.14 N·m), the normalized deviation of EL-03 was... IN-07 Normalized Deviation Torque deviation from weighting coefficient (Based on BT-02 terminal material specifications).

[0084]

[0085]

[0086] All error events in this session are summarized, and an error propagation marker matrix is ​​generated as shown in Table 4.

[0087] Table 4 Error Propagation Label Matrix

[0088] Step 5: Update the independent competency defect profile vectors of trainees in each job type based on the error propagation label matrix; ERR-IN07-04 was removed from the error set of IN-07, and ERR-IN07-03 was retained as a self-initiated error. The self-initiated errors of the two trainees were statistically analyzed according to the knowledge point dimension and the operational skill dimension, generating independent ability deficiency profile vectors as shown in Table 5.

[0089] At the same time, the proportion of responsibility contribution of EL-03 will be... IN-07 The cumulative scores are incorporated into their respective operational standardization scoring dimensions, serving as a quantitative basis for operational contributions.

[0090] Table 5. Vector of Student Independence Deficiency Profile

[0091] The signal calibration operation defect dimension of IN-07 only retains the autonomous error contribution corresponding to ERR-IN07-03. ERR-IN07-04 has been removed, and its calibration operation defect score has been significantly reduced compared to before decoupling, thus avoiding an erroneous underestimation of the capabilities of IN-07.

[0092] Step 6: Generate personalized training paths and execute joint coordination and scheduling, then synchronize them to the platform; Based on the independent capability defect profile vector in Table 5, in the practical training task knowledge graph, it is the EL-03 matching circuit wiring specification reinforcement task (T-EL-12) and PLC parameter configuration special task (T-EL-08), and the IN-07 matching signal calibration operation special task (T-IN-05).

[0093] The time windows for shared training resources involved in each candidate training path are extracted: T-EL-12 occupies the BT-02 terminal block for the 15th to 25th minute, and T-IN-05 occupies the BT-02 terminal block for the 20th to 30th minute, indicating a time conflict. A multi-agent coordination scheduling algorithm based on constraint satisfaction is used for joint solution, with the optimization objective being to maximize the rate of reduction of the average independent capability defect of the group. Hard constraints include no time conflicts in shared resources and the fulfillment of task dependencies between job types. The globally optimal path combination scheme is output: T-IN-05 is adjusted to be executed between the 27th and 37th minute to eliminate the time conflict.

[0094] Finally, the global optimal path combination scheme, the three types of error attribution ratios (EL-03: autonomous error 100%; IN-07: autonomous error 50%, degradation propagation error 50%), and the shared node degradation warning information with a 47-minute countdown to BT-02 thermal degradation are synchronized to each workstation terminal.

[0095] The data flow throughout the implementation process reflects the complete link from the original sensor data acquisition to the final path scheduling scheme: the tightening torque sensor feedback value and operation history are transformed into an equivalent contact resistance estimation sequence through the contact degradation empirical formula (step 1), and then injected into the digital twin simulation entity and quantified into a degradation impact propagation vector (step 2). The degradation impact propagation vector and the task dependency graph jointly support dual-path causal tracing (step 3). The tracing results are compared competitively to generate an error propagation label matrix (step 4). After filtering passive errors, the label matrix outputs an independent capability defect profile vector (step 5). The profile vector ultimately drives conflict-free personalized path joint scheduling (step 6). The output of each step serves as the direct input to the next step, and the data maintains causal consistency throughout the entire process.

[0096] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A smart cloud control method for multi-task collaborative training, characterized in that, include: Obtain global data from multi-task collaborative training sessions and calculate the estimated sequence of equivalent contact resistance values ​​for each shared electrical node based on the empirical formula for contact degradation. The estimated sequence of equivalent contact resistance is injected into the cloud-based digital twin simulation entity to perform signal transmission simulation and generate a degradation effect propagation vector. A dual-path causal tracing was performed on the error events of trainees in each downstream job category to obtain the correlation between human operation propagation and hardware degradation propagation. Based on the competitive comparison between the correlation between human operation propagation and the correlation between hardware degradation propagation, the error event is attributed to autonomous error, human propagation error or degradation propagation error, and an error propagation label matrix is ​​generated. Based on the error propagation marker matrix, human-caused errors and deterioration propagation errors are eliminated, generating independent capability defect profile vectors for trainees in each job category, and generating personalized training paths based on these vectors and synchronizing them to the smart cloud training platform.

2. The intelligent cloud control method for multi-task collaborative training according to claim 1, characterized in that, The global data includes the timing data of the operation behavior of trainees in each type of work, the multi-type task dependency diagram, the operation history of each shared electrical node, and the feedback value sequence of the tightening torque sensor. The empirical formula for contact degradation expresses the estimated equivalent contact resistance at the current sampling time as: the initial contact resistance value of the shared node multiplied by a degradation factor, wherein the degradation factor is calculated as follows: ； in, The degradation factor is... The degradation rate coefficient is... This represents the cumulative number of operations up to the current moment. The number of operations is the index. The torque attenuation factor is... This is the sliding average of the tightening torque feedback value up to the current moment. It is a natural constant; The degradation rate coefficient, the number of operations index, and the torque attenuation factor are empirical parameters that are pre-calibrated through experimental fitting based on the terminal material type and structural specifications. They are stored in the smart cloud training platform in the form of a parameter table according to the terminal material type and structural specifications, and are retrieved from the table when calculating the estimated sequence of equivalent contact resistance values ​​for each shared node.

3. The intelligent cloud control method for multi-task collaborative training according to claim 2, characterized in that, The step of injecting the equivalent contact resistance estimate sequence into the cloud-based digital twin simulation entity also includes: Perform thermal accumulation simulation, calculate the instantaneous heating power of each shared node based on the Joule heating formula, and perform time step iteration by combining material heat capacity parameters and heat dissipation coefficient to obtain the cumulative temperature rise curve of each shared node; The signal transmission simulation includes: based on the circuit topology of the branch where each shared node is located in the digital twin simulation entity, substituting the estimated value of the equivalent contact resistance at the current moment into the voltage division or current division equivalent model of the branch to obtain the signal attenuation at the downstream measurement point; based on the positive correlation between the thermal noise power of the contact resistance and the estimated value of the equivalent contact resistance, calculating the increment of the root mean square value of the equivalent noise voltage superimposed on the downstream signal as the noise increment. Each component of the degradation effect propagation vector corresponds to the degree of signal quality degradation of the downstream measurement channel by each shared node at each time.

4. The intelligent cloud control method for multi-task collaborative training according to claim 1, characterized in that, The first path in the dual-path causal attribution is human-driven propagation attribution, including: Tracing back along the reverse edges of the multi-task dependency graph, we find the operation outputs of upstream trainees who have a task dependency relationship with the current error event; The temporal correlation is calculated using a natural exponential function with the negative of the product of the time decay coefficient and the time interval between the upstream abnormal output time and the current error event occurrence time as the exponent. The semantic matching degree is calculated based on the set of parameter types of the upstream abnormal output and the set of input parameter types involved in the current error event, and is obtained using the Jaccard similarity coefficient algorithm. The temporal relevance and the semantic matching degree are weighted and fused to obtain the human operation propagation relevance, wherein both weight coefficients are non-negative real numbers and their sum is one.

5. The intelligent cloud control method for multi-task collaborative training according to claim 1, characterized in that, The second path in the dual-path causal attribution is the hardware degradation propagation attribution, including: Based on the degradation effect propagation vector, the downstream signal attenuation and noise increment of the shared node corresponding to the time of the error event are extracted; The direction and amplitude of the deviation of the downstream measurement point signal output by the digital twin simulation entity at this moment are compared with the direction and amplitude of the deviation of the abnormal reading actually observed by the trainee in the current error event. When the deviation directions are consistent and the amplitude ratio falls within the preset matching range, the hardware degradation propagation correlation is assigned a value of 1; otherwise, it is assigned a value of 0. The preset matching interval is a symmetrical interval centered at 1. Its upper and lower limits are determined by offline simulation calibration based on the signal modeling error range of the digital twin simulation entity for the node, and are stored in the smart cloud training platform in the form of interval endpoint pairs.

6. The intelligent cloud control method for multi-task collaborative training according to claim 1, characterized in that, When the correlation of hardware degradation propagation is higher than that of human operation propagation and exceeds a preset causal threshold, the error event is attributed to a degradation propagation error, and the source shared node that caused the degradation is marked. The preset causal threshold is a preset real number that is greater than 0 and less than 1. It is preset and stored in the smart cloud training platform according to the balance requirements of attribution misjudgment rate and missed judgment rate in the training scenario. Also includes: The torque deviation of each type of trainee in the most recent operation is normalized to the mean based on the range. Calculate the responsibility contribution ratio of each job type trainee operating the source shared node. The responsibility contribution ratio is the sum of the trainee's historical operation frequency ratio and the normalized torque deviation multiplied by the torque deviation weight coefficient, divided by the sum of the corresponding values ​​of all trainees who have operated the node. The torque deviation weighting coefficient is pre-calibrated experimentally based on the terminal material type.

7. The intelligent cloud control method for multi-task collaborative training according to claim 6, characterized in that, The generated independent competency deficit profile vectors for trainees in each trade include: The autonomous errors of trainees in each type of work are statistically analyzed according to the knowledge point dimension and the operational skill dimension involved. Each component of the independent ability defect profile vector corresponds to the degree of defect in each knowledge point dimension and operational skill dimension. The contribution of each trainee to the degradation of the shared node is included in the operational standardization scoring dimension. The operational contribution is obtained by accumulating the responsibility contribution ratio corresponding to the degradation propagation error in the error propagation label matrix.

8. The intelligent cloud control method for multi-task collaborative training according to claim 1, characterized in that, The generation of personalized training paths also includes: Based on the independent capability defect profile vector, candidate training task nodes corresponding to the defect dimension of each type of trainee are matched in the practical training task knowledge graph to generate a set of candidate training paths. Extract the shared training resource occupancy time window and cross-job task prerequisites involved in each candidate training path; Using a multi-agent coordination and scheduling algorithm based on constraint satisfaction, under the constraints of no temporal conflict in shared resources and the order of task dependencies between job types, the candidate training paths of all job types trainees are jointly solved, and the globally optimal path combination scheme that maximizes the rate of reduction of the average independent capability defect of the group as a whole within the current collaborative training session period is output to replace the set of candidate training paths. The multi-agent coordination scheduling algorithm based on constraint satisfaction takes the shared training resource occupation time window and cross-job task prerequisite dependency conditions of each candidate training path as input, takes the non-overlapping time of shared resource occupation and the satisfaction of the completion order of prerequisite tasks between jobs as hard constraints, and takes maximizing the reduction rate of the average independent ability defect profile vector of all trainees in the current collaborative training session as the optimization objective, and outputs the global optimal path combination scheme.

9. The intelligent cloud control method for multi-task collaborative training according to claim 3, characterized in that, Also includes: Perform trend extrapolation analysis on the cumulative temperature rise curves of each shared node, perform linear or polynomial extrapolation based on the assumption that the current multi-task operation mode continues, predict the estimated remaining time for each node to reach the material critical temperature, generate a thermal degradation countdown vector, and synchronize the thermal degradation countdown vector to the smart cloud training platform to trigger maintenance prompts for the shared nodes. After any trainee completes a subtask, the error propagation marker matrix, the shared resource occupancy status, and the estimated equivalent contact resistance of each node are updated in real time. This triggers dynamic re-coordination of the paths of the affected trainees and records the health decay trajectory and deterioration responsibility distribution of each shared node to the cloud-based equipment maintenance database.

10. A smart cloud control system for multi-task collaborative training, used to execute the smart cloud control method for multi-task collaborative training as described in any one of claims 1 to 9, characterized in that, include: The global data acquisition and contact resistance calculation module is used to acquire global data from multi-trade collaborative training sessions and calculate the estimated sequence of equivalent contact resistance values ​​for each shared electrical node based on the empirical formula for contact deterioration. The digital twin simulation module is used to inject the estimated sequence of equivalent contact resistance values ​​into the cloud-based digital twin simulation entity, perform signal transmission simulation, and generate a degradation effect propagation vector. The dual-path causal tracing module is used to perform dual-path causal tracing on the error events of trainees in each downstream job category, and obtain the correlation between human operation propagation and hardware degradation propagation respectively. The attribution classification module is used to attribute the error event to autonomous error, human error propagation error, or hardware degradation propagation error based on a competitive comparison between the human operation propagation correlation and the hardware degradation propagation correlation, and to generate an error propagation label matrix. The competency defect profile generation module is used to eliminate human-induced errors and deterioration-induced errors based on the error propagation marker matrix, and generate independent competency defect profile vectors for trainees in each job category. The training path generation and synchronization module is used to generate personalized training paths based on the independent capability defect profile vector and synchronize the control results to the smart cloud training platform.

Images