An intelligent practical training system based on computer neural network

Abstract

The application relates to the technical field of practical teaching, and discloses an intelligent practical training system based on a computer neural network, which comprises a knowledge explanation unit, a skill operation simulation unit and a practical training evaluation unit.The knowledge explanation unit provides individualized theoretical knowledge explanation by using a neural network model; the skill operation simulation unit constructs a simulation model to realize virtualized practice of skill operation; and the practical training evaluation unit accurately evaluates user practical training performance by using a deep learning technology.Meanwhile, the system is provided with a data acquisition unit responsible for collecting user learning behavior, operation simulation and practical training achievement data, and transmitting the data into various models to obtain output results.Finally, an intelligent feedback unit provides individualized practical training feedback and suggestions for the user according to the model output results through a defined intelligent feedback algorithm.The system integrates theoretical knowledge explanation and skill operation simulation, realizes the intelligentization and individualization of practical teaching, and effectively improves the practical teaching effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of practical training teaching, in particular to an intelligent practical training system based on computer neural networks. BACKGROUND

[0002] In the traditional practical training teaching environment, theoretical knowledge explanation and skill operation training are two core components. However, the existing teaching methods have obvious limitations in both aspects.

[0003] In terms of theoretical knowledge explanation, the traditional teaching method mainly relies on the oral transmission of teachers or written teaching materials. This method often lacks interactivity and individualization, and is difficult to adapt to the learning styles and needs of different students. Students passively accept knowledge, lack the opportunity to actively participate and think, leading to poor learning results. At the same time, teachers have heavy teaching burdens and are difficult to provide individualized explanation and guidance for each student's specific situation.

[0004] In terms of skill operation training, traditional practical training teaching needs to rely on actual equipment or simulators for practice. However, the limited number of equipment, site restrictions, and safety risks seriously restrict students' practice opportunities. Students often need to compete for equipment in limited time and cannot obtain sufficient practice and feedback. In addition, some high-risk skill operations on actual equipment have safety hazards, posing additional challenges to teaching.

[0005] With the rapid development of computer technology and artificial intelligence, neural network technology has been widely applied in various fields, providing new ideas and methods for practical training teaching. In the field of computer-aided teaching, neural network models can learn and analyze a large amount of data to provide personalized learning experience and feedback for learners. However, current intelligent practical training systems mostly focus on single theoretical knowledge explanation or skill operation simulation, lacking a comprehensive solution that combines the two organically. SUMMARY

[0006] The purpose of the present application is to provide an intelligent practical training system based on computer neural networks to solve the problems raised in the background art.

[0007] To achieve the above purpose, the present application provides the following technical solution: an intelligent practical training system based on computer neural networks, the system comprising:

[0008] a knowledge explanation unit, which constructs a knowledge explanation model, for providing theoretical knowledge explanation related to practical training;

[0009] a skill operation simulation unit, which constructs a skill operation simulation model, for simulating the skill operation process in practical training;

[0010] The practical training evaluation unit constructs a practical training evaluation model based on deep learning, which uses practical training data to evaluate the practical training performance of the user.

[0011] The data acquisition unit collects learning behavior data, operation simulation data and practical training achievement data of the user, and transmits the collected data into the knowledge explanation model, the skill operation simulation model and the practical training evaluation model respectively to obtain the output results of the models.

[0012] The intelligent feedback unit defines an intelligent feedback algorithm to analyze the output results of the models and provide personalized practical training feedback and suggestions for the user.

[0013] Preferably, the knowledge explanation model adopts a Transformer architecture, including:

[0014] Input layer: receiving practical training topics and related questions;

[0015] Encoding layer: encoding the input information through a self-attention mechanism to generate a context representation;

[0016] Decoding layer: generating corresponding explanation content according to the context representation;

[0017] Output layer: outputting theoretical knowledge explanation related to the practical training topic.

[0018] Preferably, the training steps of the knowledge explanation model include:

[0019] Collect a large number of practical training topics and corresponding knowledge explanation texts as training data;

[0020] Taking the practical training topic as input and the knowledge explanation text as label;

[0021] Using the training data to train the Transformer model, and optimizing the model by minimizing the difference between the generated text and the real text;

[0022] Using the validation set to verify the model and evaluate the explanation accuracy and fluency of the model;

[0023] When the performance of the model reaches the preset standard, save the model parameters to obtain the trained knowledge explanation model.

[0024] Preferably, the skill operation simulation model adopts a generative adversarial network (GAN) architecture, including:

[0025] Generator: receiving operation instructions to generate simulated operation process data;

[0026] Discriminator: distinguishing between generated operation process data and real operation process data;

[0027] Input layer: receiving user input operation instructions;

[0028] Output layer: output the simulated skill operation process.

[0029] Preferably, the training step of the skill operation simulation model comprises:

[0030] Collect a large amount of real skill operation process data as training data;

[0031] Take the operation instruction as the input of the generator and the real operation process data as the positive sample of the discriminator;

[0032] Train the GAN model using the training data, and optimize the model by minimizing the discrimination error of the discriminator and the difference between the generator and the real data;

[0033] Verify the model using the verification set, and evaluate the authenticity and accuracy of the simulated operation process;

[0034] When the performance of the model reaches the preset standard, save the model parameters to obtain the trained skill operation simulation model.

[0035] Preferably, the training step of the skill operation simulation model comprises:

[0036] CNN part: process the spatial features in the user's learning behavior data and operation simulation data;

[0037] LSTM part: process the time sequence features in the user's operation process;

[0038] Fully connected layer: fuse the output features of CNN and LSTM, and connect to the output layer;

[0039] Output layer: output the user's training evaluation results, including operation accuracy, completion time and step rationality.

[0040] Preferably, the training step of the skill operation simulation model comprises:

[0041] Collect the user's learning behavior data, operation simulation data and corresponding training evaluation results as training data;

[0042] Take the learning behavior data and operation simulation data as input, and take the training evaluation results as label;

[0043] Train the CNN-LSTM model using the training data, and optimize the model by minimizing the difference between the predicted evaluation results and the real evaluation results;

[0044] Verify the model using the verification set, and evaluate the evaluation accuracy and stability of the model;

[0045] When the model performance reaches the preset standard, the model parameters are saved to obtain a trained practical training evaluation model.

[0046] Preferably, the implementation steps of the intelligent feedback algorithm defined in the intelligent feedback unit include:

[0047] Step 1: define the output of the knowledge explanation model as K_Explain, the output of the skill operation simulation model as S_Simulate, and the output of the practical training evaluation model as E_Evaluate.

[0048] Step 2: according to the result of E_Evaluate, identify the weaknesses and areas for improvement of the user in the practical training;

[0049] Step 3: in combination with K_Explain, provide the user with theoretical knowledge supplementary explanation for the weaknesses;

[0050] Step 4: in combination with S_Simulate, provide the user with skill operation simulation training for the weaknesses;

[0051] Step 5: if there are multiple practical training projects or skill points, the above feedback process is performed for each practical training project or skill point respectively to obtain respective personalized feedback;

[0052] Step 6: according to the personalized feedback of all practical training projects or skill points, integrate to generate a final practical training feedback report.

[0053] Preferably, the practical training feedback report further includes a user progress tracking mechanism, and the specific steps include:

[0054] S1: record the evaluation results and feedback suggestions of the user in each practical training;

[0055] S2: compare the practical training performances of the user at different time points to analyze the progress of the user;

[0056] S3: according to the progress of the user, adjust the content of the knowledge explanation and the difficulty of the skill operation simulation to adapt to the learning pace and ability level of the user;

[0057] S4: in the practical training feedback report, show the progress trajectory of the user and the future learning suggestions.

[0058] Preferably, the data acquisition unit further includes a data enhancement module, which is responsible for data enhancement processing of the collected user learning behavior data, operation simulation data and practical training achievement data, including data expansion, data transformation and data synthesis.

[0059] Compared with the prior art, the present application has the following beneficial effects:

[0060] The knowledge explanation unit in the system builds a knowledge explanation model, which can provide theoretical knowledge explanation related to practical training using neural network technology. This way not only has rich content and various forms, but also can be personalized according to the learning style and needs of students, enhancing the interactivity and interest of learning, thereby effectively improving the learning effect of students. Through the intelligent feedback unit, the system can provide personalized practical training feedback and suggestions according to the learning behavior data and performance of students, helping students better master knowledge and skills.

[0061] The skill operation simulation unit builds a skill operation simulation model, which can simulate the skill operation process in practical training. This allows students to practice skill operation at any time and any place without relying on actual equipment or simulators, greatly breaking through the limitations of equipment quantity and site. At the same time, since the simulation operation has no safety risk, students can safely practice various high-risk skill operations, thereby obtaining more comprehensive practice opportunities and feedback.

[0062] The practical training evaluation unit builds a deep learning-based practical training evaluation model, which can objectively and accurately evaluate the practical training performance of students using practical training data. This helps teachers better understand the learning situation and progress of students, so as to develop more reasonable teaching plans and strategies.

[0063] Through the feedback of the practical training evaluation model, teachers can adjust teaching methods and content in a timely manner to improve teaching quality and effectiveness. The present invention can automatically provide knowledge explanation, skill operation simulation and practical training evaluation, greatly reducing the teaching burden of teachers. Teachers can devote more energy to teaching research and innovation to improve teaching efficiency and quality. At the same time, the system can also provide personalized teaching suggestions and guidance for teachers according to the learning data and performance of students, helping teachers better meet the learning needs of students. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 The working principle diagram of the intelligent practical training system based on computer neural network described in the present invention;

[0065] Figure 2 The training flowchart of the skill operation simulation model;

[0066] Figure 3 The network structure design diagram of the practical training evaluation model;

[0067] Figure 4 The implementation steps diagram of the intelligent feedback algorithm. DETAILED DESCRIPTION

[0068] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0069] With reference to Figures 1-4 The present application provides a technical solution: an intelligent practical training system based on computer neural networks, which comprises:

[0070] A knowledge explanation unit:

[0071] A knowledge explanation model is constructed, which is based on neural network technology and can understand and generate theoretical knowledge content related to practical training. During model training, a large amount of theoretical knowledge data, such as textbooks, teaching videos, and teaching plans, is used for learning and optimization to form a comprehensive and accurate knowledge base.

[0072] The knowledge explanation model dynamically generates personalized theoretical knowledge explanation content according to the user's learning progress and needs, supports multiple presentation forms such as text, audio, and video, and enhances the interactivity and interest of learning.

[0073] A skill operation simulation unit: a skill operation simulation model is constructed to simulate the skill operation process in practical training.

[0074] A practical training evaluation unit:

[0075] A practical training evaluation model based on deep learning is constructed, which uses practical training data (such as user operation records and achievement data) as input to evaluate the user's practical training performance.

[0076] The model learns the standards and criteria of practical training evaluation through training, can automatically identify and analyze the user's operation behavior, and gives objective and accurate evaluation results. The evaluation results include operation accuracy, efficiency, safety, and other dimensions, providing comprehensive practical training performance feedback for the user.

[0077] A data acquisition unit:

[0078] The data acquisition unit is responsible for collecting user learning behavior data, operation simulation data, and practical training achievement data. By deploying data collection modules on the user side, various data generated by users during the learning process, such as learning time, operation path, and achievement quality, are collected in real time. The collected data is preprocessed and formatted and then transmitted to the knowledge explanation model, the skill operation simulation model, and the practical training evaluation model as model input data.

[0079] An intelligent feedback unit:

[0080] An intelligent feedback algorithm is defined, which analyzes and processes the model output results to generate personalized practical training feedback and suggestions. The algorithm dynamically adjusts the feedback content and form according to the user's learning progress, operation performance and evaluation results to meet the needs of different users. Feedback content includes knowledge point prompts, operation skill suggestions, improvement directions, etc., presented to users in the form of text, audio or video to help users better master knowledge and skills.

[0081] The application will be further described below in conjunction with Examples 1 to 4:

[0082] Example 1:

[0083] The knowledge explanation model adopts a Transformer architecture, including:

[0084] Input layer: receives practical training topics and related questions as input. The practical training topic can be a specific skill name, operation process or theoretical concept, and the related question is a specific question or content that needs to be explained by the user for the practical training topic.

[0085] The input layer converts the practical training topic and related questions into numerical form that the model can process, such as word embedding vectors, for subsequent processing.

[0086] Encoding layer: uses self-attention mechanism to encode input information. Self-attention mechanism can capture the association between elements in the input sequence to generate representations containing context information. The encoding layer is stacked by multiple encoder layers, each of which contains a self-attention sublayer and a feedforward neural network sublayer, and enhances the expression ability of the model through residual connection and layer normalization.

[0087] Decoding layer: generates corresponding explanation content based on the context representation generated by the encoding layer. The decoding layer also uses self-attention mechanism, but also introduces encoder-decoder attention mechanism to capture the association between the context information output by the encoding layer and the current generated content. The decoding layer is stacked by multiple decoder layers, each of which contains a self-attention sublayer, an encoder-decoder attention sublayer and a feedforward neural network sublayer.

[0088] Output layer: the output layer converts the representation generated by the decoding layer into natural language text, i.e. theoretical knowledge explanation related to the practical training topic. The output layer usually uses softmax function to generate word probability distribution, and generates the final text through greedy search or beam search strategy.

[0089] The training steps of the knowledge explanation model include:

[0090] A large amount of practical training topics and corresponding knowledge explanation texts are collected as training data. These data can come from textbooks, teaching videos, teaching plans, and other educational resources, or be obtained through expert annotation or crowdsourcing. Ensure that the training data covers a wide range of practical training topics and knowledge points to improve the model's generalization ability.

[0091] The practical training topics are used as input, and the knowledge explanation texts are used as labels. Preprocessing operations such as word segmentation and stop word removal are performed on the input and labels to convert the text into word embedding vector form. A vocabulary is constructed, and each word in the model is assigned a unique index for use during training.

[0092] The Transformer model is trained using the training data. During training, the model parameters are optimized by minimizing the difference between the generated text and the true text. The cross-entropy loss function is used as the optimization objective, and gradient descent and other optimization algorithms are used to update the model parameters.

[0093] Set appropriate learning rates, batch sizes, and other hyperparameters to ensure the stability and convergence of the model training.

[0094] The model is validated using a validation set to evaluate the accuracy and fluency of the explanation. The validation set should contain different but related practical training topics and knowledge points from the training set. The performance of the model is evaluated by calculating the similarity (such as BLEU score) between the generated text and the true text, and manual evaluation.

[0095] When the model performance reaches the preset standard, save the model parameters to obtain the trained knowledge explanation model. Model parameters can be stored in files or databases for subsequent application.

[0096] Suppose the practical training topic is "Operation process of numerical control machine tools", and the related question is "How to set the machining parameters of numerical control machine tools?". During training, the input layer receives "Operation process of numerical control machine tools" and "How to set the machining parameters of numerical control machine tools?" as input, the encoding layer encodes the input information through self-attention mechanism to generate a representation containing context information, and the decoding layer generates the corresponding explanation content according to the context representation, such as "When setting the machining parameters of numerical control machine tools, first enter the parameter setting interface, then select the parameter type to be set, such as feed speed, spindle speed, etc., and finally input the specific value and save the settings." The output layer converts the explanation content into natural language text as the output of the model. Through training and optimization, the model can accurately and fluently answer related questions about the operation process of numerical control machine tools and provide personalized theoretical knowledge explanation for users.

[0097] Example 2:

[0098] The skill operation simulation model in the present application adopts a generative adversarial network (GAN) architecture. The architecture of the model includes:

[0099] Generator: The generator receives operation instructions as input, which can be a sequence of actions, parameter settings or other control signals input by the user through the interactive device. The generator generates simulated operation process data according to the operation instructions. These data can be the motion trajectory of the device in the virtual environment, state changes or sensor readings, etc., used to simulate the real skill operation process.

[0100] Discriminator: The task of the discriminator is to distinguish between the operation process data generated by the generator and the real operation process data. The real operation process data comes from the record in the actual skill operation process or the accurate simulation of the simulator. The discriminator compares the features of the generated data and the real data, and outputs a discrimination result, indicating the similarity between the generated data and the real data.

[0101] Input layer: The input layer is responsible for receiving the operation instructions input by the user. These instructions can be obtained through various interactive methods, such as keyboard input, mouse click, voice command or gesture recognition, etc. The input layer converts the operation instructions into numerical form that the model can process, such as vectors or matrices, for subsequent processing by the generator.

[0102] Output layer: The output layer is responsible for outputting the simulated skill operation process. This can include virtual device motion animation, state display or real-time update of sensor data, etc. The output layer converts the operation process data generated by the generator into a form that the user can perceive, so that the user can interact and practice.

[0103] The training steps of the skill operation simulation model include:

[0104] Collect a large amount of real skill operation process data as training data. These data can come from the record in the actual skill operation process, such as video, sensor data or simulator log. Ensure that the training data covers a wide range of skill operation scenarios and variants to improve the generalization ability of the model.

[0105] Use the operation instructions as the input of the generator and the real operation process data as the positive samples of the discriminator. Preprocess the input and samples as necessary, such as data cleaning, normalization or feature extraction. Construct the dataset and assign a unique identifier to each operation instruction and real operation process data in the model for use in the training process.

[0106] Train the GAN model using the training data. During the training process, optimize the model parameters by minimizing the discrimination error of the discriminator and the difference between the generator and the real data.

[0107] The goal of the discriminator is to accurately distinguish between generated data and real data, while the goal of the generator is to generate simulated data that is as close to real data as possible to deceive the discriminator.

[0108] An alternating training strategy is adopted, i.e., first training the discriminator, then training the generator, and repeating until the model reaches a stable state.

[0109] The model is verified using a validation set to evaluate the authenticity and accuracy of the simulated operation process. The validation set should contain different but related skill operation scenarios and data from the training set. The performance of the model is evaluated by comparing the similarity of the simulated operation process with the real operation process, calculating error indicators, or conducting manual evaluation.

[0110] When the model performance reaches the preset standard, save the model parameters to obtain the trained skill operation simulation model. Model parameters can be stored in files or databases for subsequent application.

[0111] Suppose the skill operation is "welding workpiece process", and the operation instructions include the movement path of the welding gun, the setting of the welding current, the control of the welding speed, etc. In the training process, the input layer receives the user's input operation instructions, such as "move the welding gun to the specified position on the workpiece, set the welding current to 100A, and weld at a speed of 5cm / s". The generator generates simulated welding process data based on the operation instructions, such as the movement trajectory of the welding gun in the virtual environment, the simulation effect of the welding arc, the temperature change of the workpiece, etc. The discriminator attempts to distinguish between these simulated data and real welding process data (such as actual welding process data recorded by high-precision sensors). Through training and optimization, the GAN model can generate more realistic welding process simulation, allowing users to practice and evaluate welding skills in a virtual environment. When the model performance reaches the preset standard, save the model parameters to obtain the trained skill operation simulation model for subsequent practical teaching.

[0112] Example 3:

[0113] The practical evaluation model adopts a convolutional neural network (CNN) combined with a long short-term memory network (LSTM) architecture. The architecture of this model includes:

[0114] CNN part: The CNN part is responsible for processing the spatial features in the user's learning behavior data and operation simulation data. Learning behavior data can include user operation records such as clicks, drags, and zooms during practical training, as well as user behavior such as perspective movement and object selection in a virtual environment. Operation simulation data can be obtained from skill operation simulation models, such as data generated by users during welding, assembly, and other operations in a virtual environment.

[0115] CNN extracts spatial features from the data, such as the position, shape, and texture of the operations.

[0116] LSTM Part: The LSTM part is responsible for processing the time series features in the user's operation process. It can capture the time order and dependency of the user's operations, such as which step the user performed first, which step the user performed later, and the time interval between these operations. LSTM can effectively handle long time series data through its unique gating mechanism, avoiding the problem of gradient vanishing or explosion.

[0117] Fully Connected Layer: The fully connected layer fuses the output features of CNN and LSTM. It concatenates or weightedly sums the spatial features extracted by CNN and the time series features extracted by LSTM to form a comprehensive feature representation. The fully connected layer maps the comprehensive features to the output layer through linear transformation and nonlinear activation function.

[0118] Output Layer: The output layer outputs the user's practical training evaluation results. The evaluation results can include operation accuracy (such as the similarity between the user's operation and the standard operation), completion time (such as the time required for the user to complete the practical training task), and step rationality (such as the logicality and orderliness of the user's operation steps).

[0119] The output layer uses a softmax function or regression model to map the comprehensive features to specific evaluation indicators.

[0120] The training steps of the practical training evaluation model include:

[0121] Collect the user's learning behavior data, operation simulation data, and corresponding practical training evaluation results as training data. The data can come from actual practical training processes, obtained through sensors, log records, or simulators.

[0122] Take the learning behavior data and operation simulation data as input, and the practical training evaluation results as label. Perform necessary preprocessing on the input data, such as data cleaning, normalization, feature extraction, etc. Construct the dataset, assign a unique identifier to each input data and label in the model, so as to use it in the training process.

[0123] Use the training data to train the CNN-LSTM model. During the training process, optimize the model parameters by minimizing the difference between the predicted evaluation results and the true evaluation results. Use cross-entropy loss function, mean square error loss function, or custom loss function to measure the difference between the predicted results and the true results.

[0124] Use gradient descent, Adam, and other optimization algorithms to update the model parameters until the model reaches a state of convergence.

[0125] The model is verified using the validation set to evaluate the evaluation accuracy and stability of the model. The validation set should contain different but related real training scenarios and user behavior data from the training set. By calculating the similarity, accuracy, recall rate and other indicators between the predicted evaluation results and the true evaluation results, the performance of the model is evaluated. At the same time, observe the performance of the model under different real training scenarios and user behaviors, and evaluate the stability and generalization ability of the model.

[0126] When the model performance reaches the preset standard, save the model parameters to obtain the trained real training evaluation model. The model parameters can be stored in a file or a database for subsequent application. At the same time, record the training process, hyperparameter settings and validation results of the model, etc. to reproduce and further optimize the model.

[0127] Suppose the real training task is "electronic component assembly", and the user's learning behavior data includes user's operation records such as selecting electronic components, moving components to designated positions, adjusting component directions, etc. in a virtual environment, and operation simulation data includes user's gesture, force, speed, etc. in the assembly process. Real training evaluation results include operation accuracy (such as whether the component assembly position is accurate), completion time (such as the time required for the user to complete the assembly task), and step rationality (such as whether the user's assembly steps meet the process requirements).

[0128] During training, learning behavior data and operation simulation data are used as input, and real training evaluation results are used as labels to train the CNN-LSTM model. The CNN part extracts spatial features of user operations, such as component position, shape, etc.; the LSTM part captures time sequence features of user operations, such as operation order, time interval, etc. The fully connected layer fuses the output features of CNN and LSTM, and connects to the output layer. The output layer outputs the user's real training evaluation results.

[0129] Through training and optimization, the CNN-LSTM model can accurately evaluate the user's performance in electronic component assembly real training, including operation accuracy, completion time and step rationality. When the model performance reaches the preset standard, save the model parameters to obtain the trained real training evaluation model, which can be used for subsequent real training teaching and evaluation.

[0130] Embodiment 4:

[0131] The implementation steps of the intelligent feedback algorithm defined by the intelligent feedback unit include:

[0132] Step 1: Define model output

[0133] Define the output of the knowledge explanation model as K_Explain, which contains the theoretical knowledge explanation content required by the user, presented in the form of text, image or video, etc.

[0134] The output of the skill operation simulation model is defined as S_Simulate, which provides simulation training for users to perform skill operations in a virtual environment, including operation steps, operation feedback, and simulation results, etc.

[0135] The output of the practical training evaluation model is defined as E_Evaluate, which contains the evaluation results of users during practical training, such as operation accuracy, completion time, step rationality, etc.

[0136] Step 2: Identify user weaknesses

[0137] According to the results of E_Evaluate, analyze the performance of users in practical training, and identify the weaknesses or areas for improvement of users in certain skill points or operation steps. For example, through the evaluation results, it is found that the user's operation accuracy is low in the assembly step of a certain electronic component, or the time required to complete a certain task is too long.

[0138] Step 3: Provide theoretical knowledge supplement explanation

[0139] Based on K_Explain, provide relevant theoretical knowledge supplement explanation for the weaknesses identified by the user. For example, for the user's weakness in electronic component assembly, provide detailed theoretical knowledge explanation about the structure, function, assembly requirements, etc. of the component.

[0140] Step 4: Provide skill operation simulation training

[0141] Based on S_Simulate, provide targeted skill operation simulation training for the weaknesses identified by the user. For example, for the user's weakness in assembly steps, provide assembly simulation training in a virtual environment, and let the user repeatedly practice until they master the correct operation skills.

[0142] Step 5: Personalized feedback process

[0143] If there are multiple practical training projects or skill points, the above feedback process is performed for each practical training project or skill point.

[0144] For each practical training project or skill point, identify weaknesses based on the results of E_Evaluate, provide theoretical knowledge supplement explanation based on K_Explain, and provide skill operation simulation training based on S_Simulate. This ensures that users can receive personalized feedback and training for each practical training project or skill point.

[0145] Step 6: Generate practical training feedback report

[0146] According to the personalized feedback of all practical training projects or skill points, the final practical training feedback report is generated. The practical training feedback report includes the user's evaluation results on each practical training project or skill point, weakness identification, theoretical knowledge supplement explanation content, skill operation simulation training situation, and improvement suggestions, etc.

[0147] In the practical training feedback report, the user progress tracking mechanism implementation steps include:

[0148] S1: In each practical training, record the user's evaluation results and feedback suggestions. The evaluation results include operation accuracy, completion time, step rationality, etc., and the feedback suggestions include theoretical knowledge supplement explanation content and skill operation simulation training suggestions, etc.

[0149] S2: Compare the user's practical training performance at different time points, analyze the user's progress in each practical training project or skill point. For example, compare the user's evaluation results before and after several training, observe the user's improvement in operation accuracy, completion time, etc.

[0150] S3: According to the user's progress, adjust the content of knowledge explanation and the difficulty of skill operation simulation.

[0151] If the user has mastered the relevant knowledge and shown good operation skills in a certain skill point, the theoretical knowledge explanation and skill operation simulation training difficulty of that skill point can be reduced.

[0152] On the contrary, if the user still has weaknesses or needs improvement in a certain skill point, the theoretical knowledge explanation and skill operation simulation training difficulty of that skill point can be increased.

[0153] S4: In the practical training feedback report, show the user's progress trajectory, including the user's evaluation result changes, weakness improvement, etc. in each practical training project or skill point. At the same time, according to the user's progress and current learning level, provide future learning suggestions for the user, such as suggesting the user to continue to strengthen the training of which skill points, learn which new theoretical knowledge, etc.

[0154] Suppose the user is doing electronic assembly training, and the practical training projects include the assembly of resistors, capacitors, and integrated circuits. In the initial training, the user's evaluation results show that the operation accuracy is low and the completion time is long in the resistor assembly steps. The intelligent feedback unit identifies that the user has weaknesses in the resistor assembly skill point according to the evaluation results.

[0155] Therefore, the intelligent feedback unit provides the user with detailed theoretical knowledge supplement explanation about the structure, function, assembly requirements, etc. of the resistor in combination with the knowledge explanation model. At the same time, in combination with the skill operation simulation model, it provides the user with virtual environment simulation training for resistor assembly, allowing the user to repeatedly practice until mastering the correct operation skills.

[0156] In the subsequent practical training, the intelligent feedback unit continues to evaluate the user's practical training performance and records the evaluation results and feedback suggestions. By comparing the user's practical training performance at different time points, the intelligent feedback unit finds that the user's operation accuracy on the resistance assembly skill point has improved, and the completion time has also been shortened.

[0157] Therefore, the intelligent feedback unit adjusts the content of knowledge explanation and the difficulty of skill operation simulation according to the user's progress. For the resistance assembly skill point, the difficulty of theoretical knowledge explanation and skill operation simulation training is reduced, and more emphasis is placed on the assembly skills of capacitors and integrated circuits.

[0158] Finally, the intelligent feedback unit integrates and generates the final practical training feedback report, showing the user's progress trajectory and future learning suggestions on each practical training project or skill point. Users can understand their learning situation and progress trajectory according to the practical training feedback report, and carry out targeted subsequent learning and training.

[0159] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0160] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A computer neural network-based intelligent practical training system, characterized in that, The system comprises: a knowledge explanation unit that builds a knowledge explanation model for providing theoretical knowledge explanation related to practical training; a skill operation simulation unit that builds a skill operation simulation model for simulating the skill operation process in practical training; a practical training evaluation unit that builds a practical training evaluation model based on deep learning, which evaluates the practical training performance of a user by using practical training data; a data acquisition unit that acquires learning behavior data, operation simulation data and practical training achievement data of the user, and transmits the acquired data into the knowledge explanation model, the skill operation simulation model and the practical training evaluation model respectively to obtain the output results of the models; an intelligent feedback unit that defines an intelligent feedback algorithm, analyzes the output results of the models and provides personalized practical training feedback and suggestions for the user; The knowledge explanation model adopts a Transformer architecture, which includes: input layer: receiving practical training topics and related questions; encoding layer: encoding the input information through a self-attention mechanism to generate context representation; decoding layer: generating corresponding explanation content according to the context representation; output layer: outputting theoretical knowledge explanation related to the practical training topic; The skill operation simulation model adopts a generative adversarial network (GAN) architecture, which includes: generator: receiving operation instructions to generate simulated operation process data; discriminator: distinguishing between generated operation process data and real operation process data; input layer: receiving user input operation instructions; output layer: outputting simulated skill operation process; The practical training evaluation model adopts an architecture combining convolutional neural network (CNN) and long short-term memory (LSTM), which includes: CNN part: processing spatial features in user learning behavior data and operation simulation data; LSTM part: processing time sequence features in user operation process; fully connected layer: fusing the output features of CNN and LSTM and connecting to the output layer; output layer: outputting the practical training evaluation results of the user, including operation accuracy, completion time and step rationality; The implementation steps of the intelligent feedback algorithm defined in the intelligent feedback unit include: Step 1: define the output of the knowledge explanation model as K_Explain, the output of the skill operation simulation model as S_Simulate, and the output of the practical training evaluation model as E_Evaluate; Step 2: according to the result of E_Evaluate, identify the weaknesses and areas for improvement of the user in practical training; Step 3: combined with K_Explain, provide the user with theoretical knowledge supplementary explanation for weaknesses; Step 4: combined with S_Simulate, provide the user with skill operation simulation training for weaknesses; Step 5: if there are multiple practical training projects or skill points, perform the above feedback process for each practical training project or skill point to obtain individualized feedback; Step 6: according to the individualized feedback of all practical training projects or skill points, integrate to generate the final practical training feedback report. 2.The intelligent practical training system based on computer neural network according to claim 1, wherein, The training steps of the knowledge explanation model include: collect a large number of practical training topics and corresponding knowledge explanation texts as training data; take the practical training topics as input and the knowledge explanation texts as labels; The Transformer model is trained using the training data by minimizing the difference between the generated text and the real text; The model is validated using the validation set to evaluate the accuracy and fluency of the explanation; When the model performance reaches the preset standard, save the model parameters to get the trained knowledge explanation model. 3.The intelligent practical training system based on computer neural network according to claim 1, wherein, The training steps of the skill operation simulation model include: Collect a large amount of real skill operation process data as training data; The operation instruction is used as the input of the generator, and the real operation process data is used as the positive sample of the discriminator; The GAN model is trained using the training data by minimizing the discrimination error of the discriminator and the difference between the generator and the real data to optimize the model; The model is validated using the validation set to evaluate the authenticity and accuracy of the simulated operation process; When the model performance reaches the preset standard, save the model parameters to get the trained skill operation simulation model.

4. The intelligent practical training system based on computer neural network according to claim 1, characterized in that, The training steps of the real training evaluation model include: Collect the user's learning behavior data, operation simulation data and corresponding real training evaluation results as training data; The learning behavior data and operation simulation data are used as input, and the real training evaluation results are used as labels; The CNN-LSTM model is trained using the training data by minimizing the difference between the predicted evaluation results and the real evaluation results to optimize the model; The model is validated using the validation set to evaluate the evaluation accuracy and stability of the model; When the model performance reaches the preset standard, save the model parameters to get the trained real training evaluation model.

5. The intelligent practical training system based on computer neural network according to claim 1, characterized in that, The real training feedback report also includes a user progress tracking mechanism, and the specific steps include: S1: Record the evaluation results and feedback suggestions of the user in each real training; S2: Compare the real training performance of the user at different time points and analyze the progress of the user; S3: According to the progress of the user, adjust the content of the knowledge explanation and the difficulty of the skill operation simulation to adapt to the learning rhythm and ability level of the user; S4: In the real training feedback report, show the progress trajectory of the user and the future learning suggestions.

6. The intelligent practical training system based on computer neural network according to claim 1, characterized in that, The data acquisition unit also includes a data enhancement module, which is responsible for data enhancement processing of the collected user learning behavior data, operation simulation data and real training achievement data, including data augmentation, data transformation and data synthesis.

Images