Centralized electrician training and checking system
Through multimodal data collection and in-depth behavioral analysis, combined with deep neural networks and adaptive question bank generation, the problems of underutilized data value and insufficient personalized services in electrician training systems have been solved, and comprehensive assessment and personalized training of trainees' operational skills and theoretical knowledge have been achieved.
Patent Information
- Application Number
- CN202510919959.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-04
- Publication Date
- 2025-09-05
AI Technical Summary
The existing electrician training system fails to fully tap the value of multimodal data during the data collection stage, operational behaviors are difficult to quantify, and personalized service capabilities are weak, especially in terms of test question generation, which makes it difficult to teach students according to their aptitude.
It adopts multimodal data acquisition, deep behavioral analysis and adaptive question bank generation technology, collects multimodal data through high-resolution cameras, microphones, and current and voltage sensors, combines deep neural networks and long short-term memory networks to build a behavioral classifier, generates personalized test questions and monitors the training environment in real time.
It achieves a comprehensive assessment of trainees' operational skills and theoretical knowledge, dynamically adjusts the difficulty and type of training content, improves the personalization and efficiency of the training process, and ensures operational safety and data integrity.
Smart Images
Figure CN120598745A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of Internet of Things, and in particular to a centralized electrician training and assessment system. Background Art
[0002] In recent years, with the rapid development of information technology, electrician training has gradually shifted from traditional classroom instruction to a modern, IT-enabled teaching model. For decades, electrician skills training primarily relied on hands-on practice and face-to-face instruction. While this model provides a direct hands-on experience, it has significant limitations in terms of resource efficiency, personalized instruction, and distance learning. Therefore, the development of a centralized electrician training system that can quantify performance and provide personalized assessments is crucial.
[0003] Most electrician training and assessment systems currently on the market still have numerous shortcomings. For example, during the data collection phase, they typically focus on a single source of information, employing crude and simplistic methods to process the collected data, failing to fully tap into the valuable insights it holds. Furthermore, many operational behaviors during electrician training are difficult to quantify, failing to accurately reflect the latest industry standards. Furthermore, the ability to provide personalized services for students of varying skill levels is limited, particularly in the generation of test questions, making it difficult to truly tailor instruction to individual needs. Summary of the Invention
[0004] In view of the above existing problems, the present invention is proposed.
[0005] Therefore, the present invention solves the problems of difficulty in quantifying operating behaviors and weak ability of personalized services during electrician training.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: The embodiment of the present invention provides a centralized electrician training and assessment system, which includes: The data preprocessing module starts the central control platform to collect multimodal data and perform preprocessing and feature extraction; The classifier construction module fuses the feature-extracted data to form a comprehensive feature vector and builds a behavior classifier on the central server; The report generation module detects the trainee's operation behavior based on the behavior classifier, generates an operation evaluation report based on the detection results, and stores it in the central database; The intelligent question bank generation module uses the central database to extract key knowledge points and skill methods, combines natural language processing methods to analyze historical test questions, and generates an intelligent question bank; The test question generation module dynamically adjusts the difficulty of the question bank and forms personalized test questions based on the students' operational performance and their mastery of electrical theory knowledge; The test question distribution module uses a central control platform to distribute test questions to trainees, applies IoT technology to monitor electrical equipment and usage in real time, and records the entire training and assessment process.
[0007] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the steps of starting the central control platform to collect multimodal data are as follows: Deploy various sensors in the training and assessment environment and start the central control server; Install a high-resolution camera in the operator's desk area and a microphone near the desk; Install high-precision current sensors and voltage sensors at key nodes of the circuit board; Through sensors, the collected multimodal data are transmitted to the central database in real time for storage.
[0008] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the preprocessing and feature extraction are specifically performed as follows: Extract multimodal data from a central database, apply a multiscale median filter to the image data in the multimodal data to remove image noise, and extract local binary pattern features in the image; Use noise reduction algorithm to remove background noise from audio data and extract Mel frequency cepstral coefficients; For the time series data of current and voltage, a low-pass filter is used to remove high-frequency noise and the moving average method is used to smooth the data and extract frequency domain features.
[0009] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the data after feature extraction is fused to form a comprehensive feature vector, and the specific steps are as follows: The multimodal data after feature extraction is normalized using z-score, and the processed normalized data is subjected to feature fusion based on information entropy and adaptive weight adjustment to form a multimodal feature vector.
[0010] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the behavior classifier is constructed on the central server, and the specific steps are as follows: The fused comprehensive feature vector is processed through the fully connected layer of the deep neural network to generate a time series feature sequence;
[0011] Input the time series feature sequence into the long short-term memory network for time series modeling; Use the softmax activation function to build a behavior classifier and generate behavior prediction results.
[0012] As a preferred solution of the centralized electrician training and assessment system described in the present invention, wherein: the trainee's operating behavior is detected by the behavior classifier, and an operation evaluation report is generated based on the detection results and stored in the central database. The specific steps are: The behavioral classifier captures the trainees' operational standardization, tool usage, and movement fluency in real time, and generates an operational evaluation report based on the captured information. Encapsulate the recorded operation assessment reports and multimodal data into a structured format, upload the data using the HTTPS encryption protocol, set automatic backup strategies and permission access, and store the assessment reports in the central database through the central controller.
[0013] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the method of extracting key knowledge points and skill methods from the central database comprises the following steps: Extract all students' operational assessment reports, multimodal data, and historical test questions from the central database and perform text cleaning and formatting; Use Chinese word segmentation tools to segment the text, and use the NER model to identify and parse the key knowledge points and skill methods in the text, and link the identified entities to the knowledge graph.
[0014] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the method of analyzing historical test questions and generating an intelligent question bank by combining natural language processing methods is specifically as follows: Use knowledge graphs to set question templates and classify history test questions into different types; Based on the different types of history test questions, and in combination with the students' learning progress and the purpose of the test, knowledge points are selected from the knowledge map as the basis for the questions; Use TF-IDF and cosine similarity to calculate the similarity between questions, recommend similar questions, and generate an intelligent question bank.
[0015] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the difficulty of the question bank is dynamically adjusted and personalized test questions are formed based on the trainees' operational performance and their mastery of electrical theoretical knowledge. The specific steps are: Collect students' actual operation data and process them using computer vision and image processing technology; Compare the processed actual operation data with the preset standard operating procedures, and evaluate the trainees' operation performance by analyzing the trainees' records in previous training and assessments; Define various question templates based on students' current learning progress and their grasp of theoretical knowledge, and automatically fill in the templates based on the parsed knowledge points and skill methods; Combine the quantitative results of students' operational performance with the evaluation of their knowledge mastery to establish personalized learning files for students; According to the comprehensive abilities of all students, clustering algorithms are used to classify and label the entire question bank; Formulate personalized test questions for students based on the classified marking results and the students' personalized learning profiles.
[0016] As a preferred solution of the centralized electrician training and assessment system described in the present invention, the central control platform distributes test questions to trainees, applies Internet of Things technology to monitor electrical equipment and usage in real time, and records the entire training and assessment process. The specific steps are: Configure the data hub by storing data information in the central database; The prepared personalized test questions are transmitted using SSL / TLS protocol and encrypted, and distributed to students through the central control platform of the data hub; The central control platform automatically activates the sensors at the workstation and starts the IoT system to monitor the equipment; Use cloud servers to detect abnormal operation, aggregate data from different sources into unified log files, and store them in the database.
[0017] The present invention offers the following beneficial effects: By integrating multimodal data acquisition, in-depth behavioral analysis, and adaptive question bank generation, it comprehensively and accurately assesses trainees' operational skills and theoretical knowledge. It also dynamically adjusts the difficulty and type of training content based on each trainee's specific performance, effectively enhancing the personalization and overall efficiency of the training process. The electrician training and assessment system utilizes the Internet of Things to monitor the training environment in real time, ensuring operational safety and data integrity, and providing strong support for cultivating more professional and practical technical personnel in the power industry. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0019] Figure 1 Schematic diagram of a centralized electrician training and assessment system in an embodiment.
[0020] Figure 2 This is a flow chart for generating personalized test questions in an embodiment. DETAILED DESCRIPTION
[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0023] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.
[0024] Reference Figure 1 and Figure 2 , is an embodiment of the present invention, which provides a centralized electrician training and assessment system, including the following steps: Start the central control platform to collect multimodal data and perform preprocessing and feature extraction.
[0025] Furthermore, various sensors are deployed in the training and assessment environment and a central control server is activated; Among them, the various sensors include a high-resolution camera with night vision function to adapt to different lighting conditions; a microphone used to capture students' voice commands and environmental sounds, and with noise reduction function to reduce the impact of background noise; and a current / voltage sensor installed at key nodes of the circuit board with high-speed sampling rate and high precision.
[0026] Install a high-resolution camera in the operator's desk area and a microphone near the desk; Install high-precision current sensors and voltage sensors at key nodes of the circuit board; Through sensors, the collected multimodal data are transmitted to the central database in real time for storage.
[0027] Configure firewalls and security protocols to protect the security of data during transmission, and set access rights to ensure that only authorized personnel can access servers and databases.
[0028] Extract multimodal data from the central database, and use a multi-scale median filter to remove image noise from the operating console image data in the multimodal data, and extract local binary pattern features in the image; The audio data is subjected to a noise reduction algorithm to remove background noise and extract Mel-frequency cepstral coefficients.
[0029] Preferably, the multi-scale median filter used gradually removes noise by applying median filters at different scales, with a total of three layers of filtering. The first layer selects a window size of 3*3 and stores the filtered image as an intermediate result. The second layer selects a window size of 5*5 and stores the filtered image as a new intermediate result. The third layer selects a larger window (7*7) and stores the filtered image as the final result. The final denoised image is saved back to the central database.
[0030] For the time series data of current and voltage, a low-pass filter is used to remove high-frequency noise and the moving average method is used to smooth the data and extract frequency domain features.
[0031] It should be noted that the smoothed current and voltage data are Fourier transformed to obtain a frequency domain representation (i.e., spectrum).
[0032] Peak detection: Detects major frequency peaks in the spectrum. Frequency peaks represent important frequency components in the data.
[0033] Energy distribution: Calculate the energy distribution of each frequency band. You can use frequency band division (for example, 0-10Hz, 10-50Hz, 50-100Hz) to analyze the energy of different frequency bands.
[0034] Spectral entropy: Calculate the entropy of the spectrum to measure the complexity and information content of the spectrum.
[0035] The feature-extracted data is fused to form a comprehensive feature vector, and a behavior classifier is built on the central server.
[0036] Furthermore, the data after feature extraction is normalized using z-score, and the normalized data is subjected to feature fusion based on information entropy and adaptive weight adjustment to form a comprehensive feature vector , the expression is, ; in, represents the comprehensive feature vector, represents the number of comprehensive features, Indicates the The weight of the comprehensive features, Indicates the The information entropy of the comprehensive features, represents the number of discretized intervals, Indicates the Comprehensive features and center points Gaussian kernel function between represents the proportional factor that balances the contributions of the two parts, represents the normalization factor, The integral function representing the frequency characteristics, The index variable representing the comprehensive feature, represents the integral variable; The expression is: ; in, represents the amplitude of the frequency characteristic, Indicates a point in time, represents the number of frequency features, Index variables representing frequency features; Among them, the information entropy of each feature is calculated , the formula is as follows: ; in, Represents comprehensive features No. The probability of a discrete value (the probability is obtained by discretizing each standardized continuous eigenvalue and then counting the frequency distribution of the frequency feature in each value interval), represents the number of discretized intervals, Index variables representing discrete values; Calculate each comprehensive feature and center point Gaussian kernel function between , the specific formula is as follows: ; in, represents the standard deviation of the Gaussian kernel; The fused comprehensive feature vector is processed by the fully connected layer of the deep neural network Processing to generate time series feature sequences ; The time series feature sequence Input into the long short-term memory network for time series modeling, the expression is: ; in, Indicates the The hidden state of time steps, Indicates the The cell state at each time step, represents the hidden state at the previous time step, represents the cell state at the previous time step, Index variable representing the time step; It should be noted that feature fusion simply combines features from different sources or different types into a long vector, which can speed up the efficiency of feature fusion.
[0037] use The activation function builds a behavior classifier and generates the Behavior prediction results for time steps , the expression is: ; in, represents the weight matrix of the classifier, Represents the bias term of the classifier; Among them, the weight matrix of the classifier Used to convert the hidden state of LSTM output Perform a linear transformation and transform the weight matrix Mapped into the classification space, and by linearly combining different hidden state features, the most helpful information for the classification task is extracted.
[0038] for The activation function constructs the behavior classifier. First, the initial weights and hyperparameters are set, and the cross entropy loss is used to If the activation function works well, it can be used to measure the difference between the predicted value and the true label. The behavior classifier is then iteratively trained using the data of the feature vector. The weights are adjusted through the backpropagation algorithm to minimize the loss function. The performance of the behavior classifier is checked regularly and, if necessary, retrained to adapt to new data or changes.
[0039] The trainee's operation behavior is detected by the behavior classifier, and an operation evaluation report is generated based on the detection results and stored in the central database.
[0040] Furthermore, the behavioral classifier can capture the trainees' operational standardization, tool usage, and movement fluency in real time, and generate an operational evaluation report based on the captured information. As each trainee performs an operation, a high-definition camera captures a video stream, while sensors collect data related to tool usage. Multimodal data is transmitted in real time to a central controller via an encrypted protocol, and then processed and analyzed by a behavior classifier. The behavior classifier focuses on three main indicators: operational standardization (e.g., whether it is performed in accordance with standard procedures), tool usage (e.g., correct selection and use of tools), and movement fluency (whether there are unnecessary pauses or repetitive movements during the operation). For each dimension, the behavior classifier will assign a score (ranging from 0 to 100 points), which is ultimately integrated into a comprehensive evaluation report.
[0041] Encapsulate recorded operation assessment reports and multimodal data into a structured format, upload data using the HTTPS encryption protocol, set automatic backup strategies and permission access, and store the operation assessment reports in a central database through a central controller; All action assessment reports generated by the behavior classifier, along with the raw multimodal data, are converted into structured JSON files and securely uploaded to a cloud server using the HTTPS protocol. An automatic backup policy is configured on the server to ensure data security, and a strict access control mechanism is in place to ensure that only authorized personnel can view specific information.
[0042] Use the central database to extract key knowledge points and skill methods, combine natural language processing methods to parse historical test questions, and generate an intelligent question bank.
[0043] Furthermore, all the students’ operational assessment reports, multimodal data, and historical test questions were extracted from the central database, and the data was cleaned and the text was formatted; Use Chinese word segmentation tools to segment the text, and use the NER model to identify and parse the key knowledge points and skill methods in the text, and link the identified entities to the knowledge graph.
[0044] Use the open source Chinese word segmentation tool (jieba) to segment the text. Then, apply a pre-trained named entity recognition (NER) model to identify key knowledge points and skills in the text. The pre-training process of the named entity recognition (NER) model is as follows: Extract all students' operational assessment reports, multimodal data, and historical examination questions from the central database as the original corpus; The original corpus is processed by the Jieba word segmentation tool to generate Chinese word sequences Load the named entity recognition (NER) model, which consists of an input layer, a context encoding layer, a label decoding layer, and an output layer. The input layer receives Chinese vocabulary sequences; Convert each Chinese word sequence into a fixed-dimensional numerical vector and add a mathematical representation of the position of the numerical vector in the sentence to each Chinese word sequence to generate a word tensor matrix containing word semantics and position information. The context encoding layer captures the deep semantic relationships and long-distance dependency features of the vocabulary tensor matrix in the sentence; The multi-layer Transformer self-attention mechanism calculates the association weight between each Chinese word sequence and the words in the sentence, generating a semantic vector of global context information; At the same time, the forward LSTM is used to process the sequence from left to right and record historical information; Use reverse LSTM to process the sequence from right to left and record future information; Concatenate historical information and future information into a complete temporal context encoding vector; The label decoding layer predicts the entity label sequence based on the context encoding vector and generates a prediction probability distribution with label transfer constraints; The output layer uses the Viterbi algorithm to decode the probability distribution results of the label decoding layer and generate the named entity recognition results (JSON format); For example, professional terms for circuit connection and welding technology are identified and transferred to the knowledge graph to achieve semantic association; Based on the identified key knowledge points and skill methods, a knowledge graph specifically targeting professional skills is constructed; The knowledge graph of professional skills consists of nodes, edges, and attributes; Among them, the identified key knowledge points (such as circuit connection) and skill methods (such as welding technology) are defined as nodes; Define the semantic relationship between knowledge points (e.g., circuit soldering belongs to electronic assembly skills) as an edge; Define the skills, methods and difficulty levels associated with each knowledge point as attributes; Integrate nodes, edges, and attributes to generate a knowledge graph of professional skills; Use knowledge graphs to set question templates and classify history test questions into different types; Based on the information in the knowledge graph, different types of question templates are designed, such as multiple-choice questions, fill-in-the-blank questions, and short-answer questions; each question type has a corresponding template, which can quickly generate specific questions.
[0045] Based on the different types of history test questions, and in combination with the students' learning progress and the purpose of the test, knowledge points are selected from the knowledge map as the basis for the questions; Use the term frequency-inverse document frequency TF-IDF algorithm and cosine similarity to calculate the similarity between questions, recommend similar questions, and generate an intelligent question bank.
[0046] Furthermore, appropriate knowledge points are selected from the knowledge graph as the basis for questions based on the student's individual learning progress and assessment objectives. The TF-IDF algorithm is used to calculate document importance, and cosine similarity is combined to measure the similarity between questions. This allows for the recommendation of similar questions to students, forming a personalized intelligent question bank.
[0047] Based on the students' operational performance and their mastery of electrical theoretical knowledge, the difficulty of the question bank is dynamically adjusted and personalized test questions are formed.
[0048] Furthermore, through the trainee operation data collected from multimodal sensors; During training, multimodal sensors are used to record students' actual operational data. This data includes video streams, audio streams, and sensor data. Trainees are also required to take regular theoretical knowledge tests to assess their understanding of electrical engineering theory.
[0049] Use computer vision technology and image processing technology to process students' actual operation data, including; Compare the processed actual operation data with the preset standard operating procedures, and evaluate the trainees' operation performance by analyzing the trainees' records in previous training and assessments; The preset standard operating procedures refer to: Step 1: Put on personal protective equipment (PPE); Step 2: Check whether the tools and equipment are in good condition; Step 3: Disconnect the power supply and perform electrical isolation; Step 4: Use a voltmeter to check whether the circuit is completely powered off; Step 5: Connect the wires according to the circuit diagram; Step 6: Reconnect the power supply and test the circuit function; Step 7: Clean the work area and document your progress.
[0050] Understand students' theoretical knowledge based on their performance in theoretical knowledge tests; Students also need to take theoretical knowledge tests regularly to assess their mastery of electrical engineering theoretical knowledge; each test contains multiple-choice questions, fill-in-the-blank questions and short-answer questions.
[0051] Define various question templates based on students' current learning progress and their grasp of theoretical knowledge, and automatically fill in the templates based on the parsed knowledge points and skill methods; First, determine the different question types based on the content and assessment requirements of electrician training. Design corresponding question templates for each question type, such as the single-choice question template: Title Description: [Title Description] Options: A. [Option A] B. [Option B] C. [Option C] D. [Option D] Correct answer: [correct answer] Scoring criteria: Full marks for correct answers, no marks for incorrect answers.
[0052] It should be noted that the question templates are automatically filled in based on the student's learning profile to generate personalized test questions. For example, if a student is weak in circuit connections, the electrician training assessment system will generate more multiple-choice questions and short-answer questions on circuit connections.
[0053] Combine the quantitative results of students' operational performance with the evaluation of their knowledge mastery to establish personalized learning files for students; Based on the students' personalized learning profiles and the comprehensive abilities of all students, the entire question bank is classified and labeled using clustering algorithms; Based on the classification marks and the student's personalized learning profile, a personalized test is prepared for each student. Each test is customized according to the student's current ability level and learning needs.
[0054] Formulate personalized test questions for students based on the classified marking results and the students' personalized learning profiles.
[0055] The central control platform distributes test questions to trainees, applies Internet of Things technology to monitor electrical equipment and usage in real time, and records the entire training and assessment process.
[0056] Furthermore, by storing data information in a central database, a data hub is configured; The prepared personalized test questions are transmitted using SSL / TLS protocol encryption, and distributed to students via the central control platform through the configuration data hub; The central control platform automatically activates the sensors at the workstation and starts the IoT system to monitor the equipment; When the trainee starts operating, the central control platform automatically activates all sensors at his or her workstation; the Internet of Things system starts to monitor the operating status of the equipment in real time, including key parameters such as temperature and current; the data collected by the sensors is transmitted to the cloud server via the wireless network.
[0057] Use cloud servers to detect abnormal operation, aggregate data from different sources into unified log files, and store them in the database; It should be noted that the cloud server is used to analyze sensor data in real time to detect whether the equipment has abnormal operating conditions, such as overheating, abnormal current, etc. Once an abnormality is detected, the electrician training and assessment system will immediately issue an alarm and record all data, including sensor readings, operation records, abnormal alarms and other information. They are all recorded in log files and stored in the database. After the training is completed, the system analyzes all collected data and generates a detailed training report including the trainees' operating performance, equipment usage and records of abnormal events.
[0058] In summary, this invention integrates multimodal data acquisition, deep behavioral analysis, and adaptive question bank generation to comprehensively and accurately assess trainees' operational skills and theoretical knowledge. It also dynamically adjusts the difficulty and type of training content based on each trainee's specific performance, effectively enhancing the personalization and overall efficiency of the training process. The electrician training and assessment system utilizes the Internet of Things to monitor the training environment in real time, ensuring operational safety and data integrity, and providing strong support for cultivating more professional and practical technical personnel in the power industry.
[0059] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A centralized electrician training and assessment system, characterized by: include, The data preprocessing module starts the central control platform to collect multimodal data and perform preprocessing and feature extraction; The classifier construction module fuses the feature-extracted data to form a comprehensive feature vector and builds a behavior classifier on the central server; The report generation module detects the trainee's operation behavior based on the behavior classifier, generates an operation evaluation report based on the detection results, and stores it in the central database; The intelligent question bank generation module uses the central database to extract key knowledge points and skill methods, combines natural language processing methods to analyze historical test questions, and generates an intelligent question bank; The test question generation module dynamically adjusts the difficulty of the question bank and forms personalized test questions based on the students' operational performance and their mastery of electrical theory knowledge; The test question distribution module uses a central control platform to distribute test questions to trainees, applies IoT technology to monitor electrical equipment and usage in real time, and records the entire training and assessment process.
2. The centralized electrician training and assessment system according to claim 1, characterized in that: The specific steps of starting the central control platform to collect multimodal data are as follows: Deploy various sensors in the training and assessment environment and start the central control server; Install a high-resolution camera in the operator's desk area and a microphone near the desk; Install high-precision current sensors and voltage sensors at key nodes of the circuit board; Through sensors, the collected multimodal data are transmitted to the central database in real time for storage.
3. The centralized electrician training and assessment system according to claim 2, characterized in that: The specific steps of preprocessing and feature extraction are as follows: Extract multimodal data from a central database, apply a multiscale median filter to the image data in the multimodal data to remove image noise, and extract local binary pattern features in the image; Use noise reduction algorithm to remove background noise from audio data and extract Mel frequency cepstral coefficients; For the time series data of current and voltage, a low-pass filter is used to remove high-frequency noise and the moving average method is used to smooth the data and extract frequency domain features.
4. The centralized electrician training and assessment system according to claim 3, characterized in that: The data after feature extraction is fused to form a comprehensive feature vector. The specific steps are: The multimodal data after feature extraction is normalized using z-score, and the processed normalized data is subjected to feature fusion based on information entropy and adaptive weight adjustment to form a multimodal feature vector.
5. The centralized electrician training and assessment system according to claim 4, characterized in that: The specific steps of building a behavior classifier on the central server are as follows: The fused comprehensive feature vector is processed through the fully connected layer of the deep neural network to generate a time series feature sequence; Input the time series feature sequence into the long short-term memory network for time series modeling; Use the softmax activation function to build a behavior classifier and generate behavior prediction results.
6. The centralized electrician training and assessment system according to claim 5, characterized in that: The specific steps of detecting the trainee's operation behavior based on the behavior classifier and generating an operation evaluation report based on the detection results and storing it in the central database are as follows: The behavioral classifier captures the trainees' operational standardization, tool usage, and movement fluency in real time, and generates an operational evaluation report based on the captured information. Encapsulate the recorded operation assessment reports and multimodal data into a structured format, upload the data using the HTTPS encryption protocol, set automatic backup strategies and permission access, and store the assessment reports in the central database through the central controller.
7. The centralized electrician training and assessment system according to claim 6, characterized in that: The method of extracting key knowledge points and skills from the central database includes the following specific steps: Extract all students' operational assessment reports, multimodal data, and historical test questions from the central database and perform text cleaning and formatting; Use Chinese word segmentation tools to segment the text, and use the NER model to identify and parse the key knowledge points and skill methods in the text, and link the identified entities to the knowledge graph.
8. The centralized electrician training and assessment system according to claim 7, characterized in that: The specific steps of combining natural language processing method to analyze historical test questions and generate intelligent question bank are as follows: Use knowledge graphs to set question templates and classify history test questions into different types; Based on the different types of history test questions, and in combination with the students' learning progress and the purpose of the test, knowledge points are selected from the knowledge map as the basis for the questions; Use TF-IDF and cosine similarity to calculate the similarity between questions, recommend similar questions, and generate an intelligent question bank.
9. The centralized electrician training and assessment system according to claim 8, characterized in that: The said method dynamically adjusts the difficulty of the question bank and forms personalized test questions based on the students' operation performance and their mastery of electrical theoretical knowledge. The specific steps are: Collect students' actual operation data and process them using computer vision and image processing technology; Compare the processed actual operation data with the preset standard operating procedures, and evaluate the trainees' operation performance by analyzing the trainees' records in previous training and assessments; Define various question templates based on students' current learning progress and their grasp of theoretical knowledge, and automatically fill in the templates based on the parsed knowledge points and skill methods; Combine the quantitative results of students' operational performance with the evaluation of their knowledge mastery to establish personalized learning files for students; According to the comprehensive abilities of all students, clustering algorithms are used to classify and label the entire question bank; Formulate personalized test questions for students based on the classified marking results and the students' personalized learning profiles.
10. The centralized electrician training and assessment system according to claim 9, characterized in that: The central control platform distributes test questions to trainees, uses Internet of Things technology to monitor electrical equipment and usage in real time, and records the entire training and assessment process. The specific steps are: Configure the data hub by storing data information in the central database; The prepared personalized test questions are transmitted using SSL / TLS protocol and encrypted, and distributed to students through the central control platform of the data hub; The central control platform automatically activates the sensors at the workstation and starts the IoT system to monitor the equipment; Use cloud servers to detect abnormal operation, aggregate data from different sources into unified log files, and store them in the database.