Proprietary education fusion practical training effect simulation and optimization method based on digital twinborn body

Abstract

According to the digital twinborn fusion practical training effect simulation and optimization method provided by the invention, an innovative multi-agent simulation model can simulate a nonlinear growth track of student ability under different teaching parameters in the digital twinborn, so that teaching optimization is changed from an empirical mode of relying on post-evaluation to an empirical mode of relying on post-evaluation to an empirical mode of relying on post-evaluation to an empirical mode of relying on post-evaluation to an empirical mode of relying on post-evaluation. And converting into a scientific decision-making mode based on prediction simulation. By combining advanced algorithms such as Bayesian optimization and SHAP contribution degree analysis, the system not only can automatically search for a globally optimal teaching parameter combination, but also can quantitatively explain the specific influence of each parameter on the effect, thereby outputting a parameter adjustment suggestion report with scientificity, operability and interpretability for teachers. The method can dynamically adapt to individual differences and teaching requirements, remarkably improves the accuracy, adaptability and overall effect of practical teaching, and effectively solves the core problems of virtual-real disjunction, evaluation lag, path solidification and the like in the prior art.

Description

Technical Field

[0001] This invention relates to the field of practical training technology, and in particular to a method for simulating and optimizing the effectiveness of industry-education integration practical training based on digital twins. Background Technology

[0002] In recent years, deepening the integration of industry and education and improving the quality of practical teaching have become important directions for higher education reform in my country. University practical training, as a crucial link connecting theoretical teaching and industrial practice, directly affects the quality of cultivating application-oriented talents. Traditional practical training models, relying mainly on physical laboratories, fixed processes, and experienced teacher guidance, are no longer sufficient to meet the high demands of modern industries for talents' rapid adaptability, complex problem-solving abilities, and innovation capabilities. To improve the effectiveness of practical training, various information technology and virtualization methods have been introduced into teaching, such as using virtual reality (VR) technology to construct simulated operating environments or developing online training platforms to record students' operating steps. These technologies have, to some extent, enriched teaching methods and alleviated some of the pressure on hardware resources.

[0003] However, existing practical training models and evaluation systems still have several significant problems, hindering the depth and effectiveness of industry-education integration. First, there is a disconnect between virtual and real-world components, lacking dynamic mapping and feedback. Existing virtual simulation systems are mostly static simulations with pre-set scripts, unable to link with the real physical equipment in real time. Students practice in a closed virtual environment, but their operations cannot realistically drive or reflect the operation of physical equipment, resulting in a disconnect between the virtual and the real. The practical training experience differs from real industrial scenarios, making it difficult to cultivate students' genuine engineering skills. Second, teaching evaluation is lagging and one-sided, lacking predictive and optimization capabilities. Existing technologies mostly focus on post-event evaluation of student performance (as described in CN 120298180 A, using multi-dimensional assessment through analysis of operation trajectories, voice, and other data). Although the evaluation dimensions are more comprehensive, it is essentially a "descriptive" analysis, only informing teachers and students "how they performed in the past." It cannot perform "predictive" simulations of the potential future effects of different teaching strategies (such as adjusting task difficulty, changing instruction methods, and reorganizing resources) before or during practical training, thus failing to provide forward-looking decision support for the dynamic optimization of teaching pathways. Furthermore, the training pathways are rigid and lack personalized adaptability. Teaching parameters (such as task sequences and instruction intensity) are usually set uniformly by teachers based on experience, lacking a data-driven adaptive adjustment mechanism tailored to individual student abilities, making it difficult to achieve "teaching according to aptitude," resulting in some students being "under-challenged" or "falling behind."

[0004] These problems have led to widespread difficulties in practical training in universities, including poor effectiveness, inefficient resource utilization, and a need to improve the match between talent cultivation and industry needs. Specific impacts or drawbacks include: a disconnect between practical training content and real-world industry dynamics, resulting in insufficient development of students' ability to solve complex engineering problems; reliance on teacher experience for teaching adjustments, lacking scientific basis, leading to slow optimization processes and high trial-and-error costs; and difficulty in implementing personalized teaching on a large scale, hindering the precise cultivation of highly skilled technical personnel.

[0005] Therefore, the urgent technical problem to be solved in this field is: how to construct a new method that can deeply integrate virtual and real environments and realize forward-looking simulation and dynamic optimization decision support for the training process, so as to break through the bottlenecks of existing training models in terms of real-time performance, predictability and personalization, and thus scientifically and efficiently improve the overall effectiveness of industry-education integration training. Summary of the Invention

[0006] The purpose of this invention is to provide a method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins, so as to solve the problems existing in the prior art.

[0007] To achieve the above objectives, the present invention provides the following solution: This invention provides a method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins, including: S1. Collect the operating status data of physical equipment in the physical training environment and the multimodal operation behavior data of students, and synchronize the operating status data and multimodal operation behavior data to the virtual space in real time to construct a dynamic digital twin corresponding to the physical training environment; S2. In the dynamic digital twin, a student ability growth simulation model is established based on historical and real-time data. The student ability growth simulation model is used to simulate the evolution of students' ability status under different combinations of teaching parameters. S3. Define an objective function with the goal of comprehensive training effectiveness, and based on the student ability growth simulation model, use a parameter optimization algorithm to iteratively optimize the combination of teaching parameters to find the recommended combination of teaching parameters that makes the objective function value optimal. S4. Based on simulation data under the recommended combination of teaching parameters, the contribution measurement algorithm is used to analyze the marginal contribution of each teaching parameter to the training effectiveness, and a visual analysis report containing parameter adjustment suggestions is generated.

[0008] Preferably, step S1 specifically includes: S11. Collect time-series data of the equipment's operating status through IoT sensors deployed on physical training equipment; collect students' operation sequences, trajectories, time consumption, and physiological index data through motion capture equipment, operation log systems, and physiological sensors as multimodal operation behavior data; S12. Use edge computing nodes to preprocess and align the collected runtime status time-series data and multimodal operation behavior data with timestamps, and transmit the synchronized data stream to the twin data engine through message middleware; S13. The twin data engine integrates the received data stream with the high-fidelity 3D model, drives the device model and virtual student avatar status update in the virtual environment, realizes the dynamic mapping of all elements from the physical environment to the virtual environment, and completes the construction and synchronization of the dynamic digital twin.

[0009] Preferably, in step S12, the preprocessing includes: using Kalman filtering to denoise and smooth the state estimation of the sensor data; and using an ontology-based data fusion method to establish a unified semantic mapping for multi-source data, ensuring that the operating state data and the operation behavior data are logically related.

[0010] Preferably, in step S2, the student ability development simulation model is a multi-agent model built based on a reinforcement learning framework, and its modeling method includes: Each student is defined as an intelligent agent, whose state is represented as a multi-dimensional vector of knowledge level, skill proficiency, and fatigue. Define the action space as the set of operations that students can perform; Define the reward function, with the following formula: ; Used to quantify the immediate results of a single step; The dynamic process of the model is described by the state transition function, and the function formula is: ; in, for The student's ability state vector at time t. , For knowledge level, For skill proficiency, For fatigue level, For the current combination of teaching parameters, for The actions students perform at all times include task difficulty, frequency of guidance and intervention, and resource allocation; Monte Carlo simulations are performed using a model to predict, given parameters, the future starting from the current state. The expected capability gain after the step is given by the formula: ; in, For expectation operator, For the current time step, To predict the step size, Strategies for student intelligent agents As a discount factor, For the first Instant rewards for each step This is a vector of teaching parameters.

[0011] Preferably, step S3 specifically includes: S31. Define the objective function: ; in, For the first The student or the first The weight of each capability dimension The ability gain predicted by the student ability growth simulation model; S32. Bayesian optimization is used as the parameter optimization algorithm, and Gaussian process is used as a surrogate model to fit the black-box relationship between the objective function and the parameters; S33. Improve the acquisition function by maximizing the expected value, and iteratively select the next combination of parameters to be simulated; S34. Input the next combination of parameters to be simulated into the student ability growth simulation model for simulation, obtain the corresponding objective function estimate, and update the surrogate model; S35. Repeat steps S33 and S34 until the convergence condition is met, and output the recommended combination of teaching parameters that maximizes the objective function.

[0012] Preferably, in step S4, the contribution measurement algorithm uses SHAP value analysis, and the specific steps include: S41. Collect multiple sets of simulation data generated during the parameter optimization process and construct a dataset; S42. Train an effectiveness prediction model on the dataset using the gradient boosting decision tree algorithm; S43. For the recommended combination of teaching parameters and its predicted effectiveness, calculate the SHAP value of each teaching parameter component; the SHAP value is approximated by the following formula: ; in, Index of specific teaching parameters for the contribution to be evaluated. For the set of all teaching parameters, For the total number of parameters, For a subset that does not contain parameters, To predict the expected value for a model that takes only parameters from a subset as input. The marginal contribution of the parameter to the final prediction performance relative to the average level. To predict the expected value of the model after adding parameters to a subset, These are the weighting coefficients.

[0013] Preferably, in step S4, generating the visualization analysis report specifically includes: Integrate recommended combinations of teaching parameters, predicted optimal outcomes, and SHAP values ​​for each parameter; Generate reports for teachers, including: suggestions for optimal teaching parameter configurations for the class as a whole, ranking and explanation of the influence of each parameter based on SHAP values, and warnings for differentiated teaching for students with abnormal performance. Generate reports for students, including: personalized learning path recommendations based on their personal historical data simulation, radar charts of their current ability dimensions, and expected growth curves under the recommended path.

[0014] This invention also provides a simulation and optimization system for the effectiveness of industry-education integration training based on digital twins, comprising: The twin synchronization construction module is used to collect data from the physical training environment and construct a dynamic digital twin; The performance simulation and prediction module is used to run a student ability growth simulation model in a dynamic digital twin to predict the student ability growth trajectory under different teaching parameters. The teaching parameter optimization module is used to call the effectiveness simulation and prediction module to perform multiple simulations with the goal of comprehensive training effectiveness, and to search for the optimal combination of teaching parameters through the parameter optimization algorithm. The contribution analysis and report generation module is used to analyze the contribution of each teaching parameter and generate a visual analysis report.

[0015] Preferably, the twin synchronization construction module further includes: A multi-source data acquisition unit is used to collect data through IoT sensors, motion capture devices, and logging systems. The edge computing and synchronization unit is used to perform data preprocessing and alignment at the edge and synchronize data in real time via a message queue. The 3D twin engine unit is used to integrate high-fidelity models with real-time data streams to drive dynamic updates of the virtual environment.

[0016] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the aforementioned method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins.

[0017] The present invention achieves the following beneficial technical effects compared to the prior art: This invention provides a method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins. This method completely breaks down the barrier between virtual simulation and real operation by constructing a dynamic digital twin that is synchronized in real time with all elements of the physical training environment. It achieves a high-fidelity mapping of the training process in virtual space, laying a data foundation for in-depth analysis. Furthermore, its innovative multi-agent simulation model can simulate the nonlinear growth trajectory of students' abilities under different teaching parameters within the digital twin, transforming teaching optimization from an experience-based model relying on ex-post evaluation to a scientific decision-making model based on predictive simulation. By combining advanced algorithms such as Bayesian optimization and SHAP contribution analysis, the system can not only automatically search for the globally optimal combination of teaching parameters but also quantify and explain the specific impact of each parameter on effectiveness, thus providing teachers with parameter adjustment suggestion reports that are scientific, operable, and interpretable. Ultimately, this invention forms a complete closed-loop optimization system from "real-time perception - dynamic twin - predictive simulation - intelligent optimization - attribution feedback", which can dynamically adapt to individual differences and teaching needs, significantly improve the accuracy, adaptability and overall effectiveness of practical training, and effectively solve the core problems in the existing technology such as the disconnect between virtual and reality, lagging evaluation and rigid path. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 The flowchart of the simulation and optimization method for the effectiveness of industry-education integration training based on digital twins provided by this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The purpose of this invention is to provide a method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins, so as to solve the problems existing in the prior art.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] Example 1: Figure 1 This paper demonstrates the overall process of the industry-education integration training effectiveness simulation and optimization method based on digital twins provided by this invention. The core of this invention lies in constructing a dynamic digital twin that is synchronized in real time with the physical training environment, and performing teaching effectiveness simulation and optimization based on intelligent algorithms in this virtual space.

[0024] After startup, the entire system first enters the digital twin synchronous construction phase. During this phase, the system collects real-time operational status data, including joint position, velocity, torque, end effector pose, and program execution status, from various IoT sensors (such as current sensors, encoders, six-dimensional force sensors, and vision cameras) deployed on the industrial robot body, controller, and work units. Simultaneously, multimodal operational behavior data of students is collected through motion capture gloves, eye trackers, smart bracelets, and the operation log interface of the training platform. This data includes hand movement trajectories, gaze focus sequences, program code editing history, button click streams, and physiological indicators such as heart rate and skin conductance. These heterogeneous data streams are sent to edge computing nodes deployed in the training lab. The edge nodes first timestamp-align the raw data from each sensor. For noisy sensor data (such as torque signals), a Kalman filter algorithm is used for denoising and state smoothing estimation. The state equation and observation equation are established based on the physical kinematic model of the device.

[0025] Subsequently, the system employs an ontology-based data fusion method to establish a unified semantic mapping for data from different sources. For example, it maps "the real-time angle of robot joint 1 (from the encoder)," "the target angle of joint 1 set in the student program (from the log)," and "the rotation of joint 1 in the virtual model (from the 3D engine)" to the unified concept of "robot joint 1 state," ensuring a precise logical association between the physical entity, student operations, and the virtual model. The preprocessed and fused data is transmitted in real-time to the twin data engine on the central server via message middleware such as Apache Kafka, achieving high throughput and low latency. This engine integrates a high-fidelity 3D model of the industrial robot and its working environment. Upon receiving the data stream, it drives the robot model in the virtual scene to perform movements completely consistent with the physical entity in real time, and updates the posture and operational state of the virtual student avatar, thereby completing the construction and continuous updating of a digital twin that is dynamically synchronized with the physical training environment in all elements and states.

[0026] After the dynamic digital twin has been running stably and accumulated a certain amount of historical data, the system enters the student ability development simulation modeling stage. This stage aims to predict student ability development under different teaching strategies within the virtual "sandbox" provided by the twin. This invention employs a multi-agent model based on a reinforcement learning framework to achieve this goal. The model abstracts each student participating in the training as an independent agent. Each agent... state of time Represented by a multidimensional vector, for example ,in, It represents the level of theoretical knowledge (which can be quantified based on theoretical test scores and historical Q&A data). Represents skill proficiency (calculated based on historical success rate, efficiency, and standardization of operations). This represents fatigue or cognitive load (inferred from physiological sensor data and operational performance trends). The agent's action space is defined as the set of operations that students can perform during training, such as "writing a linear motion instruction," "adjusting motion speed parameters," "manually teaching a path point," and "requesting to view help documentation." To quantify the immediate effectiveness of a single-step operation, the system defines a reward function. The dynamic process of student ability evolution is described by a state transition function, the formula of which is: ; in, for The student's ability state vector at time t. , For knowledge level, For skill proficiency, For fatigue level, For the current combination of teaching parameters, such as , This represents the difficulty of the task (such as path complexity and accuracy requirements). This indicates the frequency of guidance and intervention (such as the frequency and level of detail of system prompts). This represents resource allocation (such as the accessibility of virtual assistive tools). Each student agent's initial policy is trained using its historical operational data. During simulation prediction, for a given set of teaching parameters to be evaluated, the system runs a Monte Carlo simulation in a digital twin, simulating multiple learning trajectories of the agent in this new environment, thereby predicting future learning paths starting from the current state. The expected capability gain after one step (e.g., the next 5 practical training sessions) is calculated using the following formula: ; in, For expectation operator, For the current time step, To predict the step size, Strategies for student intelligent agents As a discount factor, For the first Instant rewards for each step This is a vector of teaching parameters. Through this model, teachers can "see" in advance the potential impact of adjusting task difficulty or instructional methods on students' long-term competence development without actually implementing the changes.

[0027] Based on simulation prediction capabilities, the system enters the intelligent optimization stage of teaching parameters. The goal of this stage is to automatically find the optimal combination of teaching parameters that maximizes the overall or specific group's training effectiveness. First, the system defines a comprehensive training effectiveness objective function. For example, for optimizing a class, the objective function can be defined as: ; in, For the first The student or the first The weight of each capability dimension This invention predicts the ability gain for a student ability development simulation model. Due to the high computational cost of simulation models and the complex black-box nature of the relationship between the objective function and parameters, this invention employs Bayesian optimization as the parameter optimization algorithm. Bayesian optimization uses a Gaussian process as a surrogate model to fit a small amount of existing sample data and provides the probability distribution (mean and variance) of the objective function across the entire parameter space. Then, by maximizing the expected value to improve the sampling function, the algorithm intelligently selects the next most promising parameter combination for simulation evaluation. After inputting the parameter combination into the student ability development simulation model to obtain the estimated value of the next objective function, the Gaussian process surrogate model is updated using this new sample. This "evaluation-update-selection" iterative cycle continues until the preset number of iterations or convergence conditions are reached, ultimately outputting a recommended combination of teaching parameters that maximizes (or satisfies) the objective function. This process navigates the vast parameter space efficiently, avoiding blind grid searches or random trials.

[0028] After obtaining the optimal parameter combination and its corresponding predicted performance, the system enters the contribution analysis and report generation stage. To make the optimization decision more interpretable, this invention uses SHAP values ​​for contribution analysis. The system collects all simulation data generated during the Bayesian optimization process to construct a dataset. An interpretable performance prediction model is trained on this dataset using the gradient boosting decision tree algorithm. For the finally recommended parameters, the SHAP value of each parameter component is calculated. The SHAP value is based on cooperative game theory, fairly allocating the contribution of each feature in the total performance, and its calculation formula is: ; in, Index of specific teaching parameters for the contribution to be evaluated. For the set of all teaching parameters, For the total number of parameters, For a subset that does not contain parameters, To predict the expected value for a model that takes only parameters from a subset as input. The marginal contribution of the parameter to the final prediction performance relative to the average level. To predict the expected value of the model after adding parameters to a subset, The weighting coefficients are used. Finally, the system integrates the recommended combination of teaching parameters, the predicted optimal results, and the SHAP values ​​of each parameter, automatically generating a visual analysis report. The teacher-side report displays the optimal parameter configuration suggestions for the entire class, clearly presenting the SHAP values ​​of each teaching parameter in the form of bar charts, explicitly informing teachers that "increasing the difficulty of tasks has the greatest positive contribution to skill development at the current stage, but has a slight negative impact on the confidence of some students," while also marking students who are not suited to the general approach and require individual attention. Students receive personalized learning path maps, including expected growth curves and ability radar charts simulated based on their own historical data under the recommended parameters, providing them with clear learning goals and path guidance.

[0029] In summary, this invention, through the close connection and coordinated operation of the four stages mentioned above, achieves a complete closed loop, from real-time perception and twinning of the physical training environment to prediction, simulation, and parameter optimization of teaching effectiveness in virtual space, ultimately outputting interpretable and accurate optimization suggestions. This method deeply integrates digital twins, reinforcement learning, Bayesian optimization, and interpretable artificial intelligence technologies, providing a systematic technical solution to overcome the bottlenecks of traditional practical training in terms of personalization, forward-looking perspectives, and scientific rigor.

[0030] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0031] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0032] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins, characterized in that, include: S1. Collect the operating status data of physical equipment in the physical training environment and the multimodal operation behavior data of students, and synchronize the operating status data and multimodal operation behavior data to the virtual space in real time to construct a dynamic digital twin corresponding to the physical training environment; S2. In the dynamic digital twin, a student ability growth simulation model is established based on historical and real-time data. The student ability growth simulation model is used to simulate the evolution of students' ability status under different combinations of teaching parameters. S3. Define an objective function with the goal of comprehensive training effectiveness, and based on the student ability growth simulation model, use a parameter optimization algorithm to iteratively optimize the combination of teaching parameters to find the recommended combination of teaching parameters that makes the objective function value optimal. S4. Based on simulation data under the recommended combination of teaching parameters, the contribution measurement algorithm is used to analyze the marginal contribution of each teaching parameter to the training effectiveness, and a visual analysis report containing parameter adjustment suggestions is generated.

2. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 1, characterized in that, Step S1 specifically includes: S11. Collect time-series data of the equipment's operating status through IoT sensors deployed on physical training equipment; collect students' operation sequences, trajectories, time consumption, and physiological index data through motion capture equipment, operation log systems, and physiological sensors as multimodal operation behavior data; S12. Use edge computing nodes to preprocess and align the collected runtime status time-series data and multimodal operation behavior data with timestamps, and transmit the synchronized data stream to the twin data engine through message middleware; S13. The twin data engine integrates the received data stream with the high-fidelity 3D model, drives the device model and virtual student avatar status update in the virtual environment, realizes the dynamic mapping of all elements from the physical environment to the virtual environment, and completes the construction and synchronization of the dynamic digital twin.

3. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 2, characterized in that, In step S12, the preprocessing includes: using Kalman filtering to denoise and smooth the state estimation of sensor data; and using an ontology-based data fusion method to establish a unified semantic mapping for multi-source data, ensuring that the operating state data and the operation behavior data are logically related.

4. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 1, characterized in that, In step S2, the student ability development simulation model is a multi-agent model built based on a reinforcement learning framework, and its modeling method includes: Each student is defined as an intelligent agent, whose state is represented as a multi-dimensional vector of knowledge level, skill proficiency, and fatigue. Define the action space as the set of operations that students can perform; Define the reward function, with the following formula: ； Used to quantify the immediate results of a single step; The dynamic process of the model is described by the state transition function, and the function formula is: ； in, for The student's ability state vector at time t. , For knowledge level, For skill proficiency, For fatigue level, For the current combination of teaching parameters, for The actions students perform at all times include task difficulty, frequency of guidance and intervention, and resource allocation; Monte Carlo simulations are performed using a model to predict, given parameters, the future starting from the current state. The expected capability gain after the step is given by the formula: ； in, For expectation operator, For the current time step, To predict the step size, Strategies for student intelligent agents As a discount factor, For the first Instant rewards for each step This is a vector of teaching parameters.

5. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 4, characterized in that, Step S3 specifically includes: S31. Define the objective function: ； in, For the first The student or the first The weight of each capability dimension The ability gain predicted by the student ability growth simulation model; S32. Bayesian optimization is used as the parameter optimization algorithm, and Gaussian process is used as a surrogate model to fit the black-box relationship between the objective function and the parameters; S33. Improve the acquisition function by maximizing the expected value, and iteratively select the next combination of parameters to be simulated; S34. Input the next combination of parameters to be simulated into the student ability growth simulation model for simulation, obtain the corresponding objective function estimate, and update the surrogate model; S35. Repeat steps S33 and S34 until the convergence condition is met, and output the recommended combination of teaching parameters that maximizes the objective function.

6. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 1, characterized in that, In step S4, the contribution measurement algorithm uses SHAP value analysis, and the specific steps include: S41. Collect multiple sets of simulation data generated during the parameter optimization process and construct a dataset; S42. Train an effectiveness prediction model on the dataset using the gradient boosting decision tree algorithm; S43. For the recommended combination of teaching parameters and its predicted effectiveness, calculate the SHAP value of each teaching parameter component; the SHAP value is approximated by the following formula: ； in, Index of specific teaching parameters for the contribution to be evaluated. For the set of all teaching parameters, For the total number of parameters, For a subset that does not contain parameters, To predict the expected value for a model that takes only parameters from a subset as input. The marginal contribution of the parameter to the final prediction performance relative to the average level. To predict the expected value of the model after adding parameters to a subset, These are the weighting coefficients.

7. The method for simulating and optimizing the effectiveness of industry-education integration training based on digital twins according to claim 6, characterized in that, Step S4, generating the visualization analysis report specifically includes: Integrate recommended combinations of teaching parameters, predicted optimal outcomes, and SHAP values ​​for each parameter; Generate reports for teachers, including: suggestions for optimal teaching parameter configurations for the class as a whole, ranking and explanation of the influence of each parameter based on SHAP values, and warnings for differentiated teaching for students with abnormal performance. Generate reports for students, including: personalized learning path recommendations based on their personal historical data simulation, radar charts of their current ability dimensions, and expected growth curves under the recommended path.

8. A simulation and optimization system for the effectiveness of industry-education integration training based on digital twins, characterized in that: include: The twin synchronization construction module is used to collect data from the physical training environment and construct a dynamic digital twin; The performance simulation and prediction module is used to run a student ability growth simulation model in a dynamic digital twin to predict the student ability growth trajectory under different teaching parameters. The teaching parameter optimization module is used to call the effectiveness simulation and prediction module to perform multiple simulations with the goal of comprehensive training effectiveness, and to search for the optimal combination of teaching parameters through the parameter optimization algorithm. The contribution analysis and report generation module is used to analyze the contribution of each teaching parameter and generate a visual analysis report.

9. The simulation and optimization system for industry-education integration training effectiveness based on digital twins according to claim 8, characterized in that, The twin synchronization construction module further includes: A multi-source data acquisition unit is used to collect data through IoT sensors, motion capture devices, and logging systems. The edge computing and synchronization unit is used to perform data preprocessing and alignment at the edge and synchronize data in real time via a message queue. The 3D twin engine unit is used to integrate high-fidelity models with real-time data streams to drive dynamic updates of the virtual environment.

10. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the simulation and optimization method for the effectiveness of industry-education integration training based on digital twins as described in any one of claims 1 to 7.

Images