AI Adaptive Training System and Method for Safety Standardization Information of Nonferrous Heavy Metals

Abstract

This invention discloses an AI-adaptive training system and method for non-ferrous heavy metal safety standardization information, belonging to the field of enterprise safety production information management technology. Based on enterprise safety procedures and training materials, this method constructs a data-driven closed-loop management system through five steps: training planning, examination and assessment, personnel status monitoring, personalized training, and needs analysis and feedback. This invention utilizes natural language processing technology to parse documents and generate test questions and courseware; it uses semantic similarity algorithms to automatically grade subjective questions, identifying employee knowledge weaknesses and historical score rates; it employs image recognition for online training identity verification; and it uses data clustering technology to analyze common knowledge gaps, generating individual and departmental training needs plans, which are then fed back into the formulation of the next year's training plan. This invention solves the problems of insufficient targeting and low management efficiency in traditional training, significantly improving the accuracy of training and the effectiveness of risk management.

Description

Technical Field

[0001] This invention relates to the field of enterprise safety production information management technology, specifically to an AI adaptive training system and method for non-ferrous heavy metal safety standardization information. Background Technology

[0002] In the production process of an enterprise, safety training and education are the core links of safety management, directly related to employees' safety awareness and operational compliance. Traditional safety training in non-ferrous heavy metal enterprises often adopts a model of "long-term intensive learning + final examination and acceptance," which has revealed many limitations in practice.

[0003] First, the training is ineffective. Heavy production tasks conflict with concentrated training sessions, making it difficult for employees to dedicate sufficient energy. This can even negatively impact their work performance, creating safety hazards. Second, training management is inefficient. Safety managers must manually compile large amounts of training plans, exam papers, and grade records, a tedious task prone to errors in grading, statistics, and inaccurate recording. This makes them particularly vulnerable during safety standardization inspections. Furthermore, frontline employees resist training that takes up work or rest time, resulting in low participation. Delayed information transmission during personnel transfers, return to work, and certificate renewals can easily lead to training omissions and blind spots in safety management. While some online training platforms exist, most are limited in function, merely providing digital displays of training content or simple online exams. They fail to fundamentally address deeper issues such as the mismatch between training content and individual capabilities, fragmented management processes, and inadequate personalized instruction.

[0004] Therefore, existing technologies suffer from problems such as insufficient accuracy and effectiveness in training, low management efficiency, low participation enthusiasm, and insufficient management coverage. Summary of the Invention

[0005] The purpose of this invention is to provide an AI-adaptive training system and method for non-ferrous heavy metal safety standardization information, so as to solve the above-mentioned problems existing in the prior art, and to deeply integrate with the enterprise safety management system and realize intelligent and personalized training throughout the entire process.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An AI-adaptive training method for safety standardization information in non-ferrous heavy metals, the method being based on the safety regulations and training materials of non-ferrous heavy metal enterprises, and constructing a data-driven closed-loop training management system through the following steps: Training plan steps: Based on the three-level organizational structure of company, factory and work team, and according to the preset approval path rules corresponding to the three-level structure, an electronic training plan is generated. The plan is reviewed and released through the corresponding multi-level electronic approval process, and reminder messages are sent to relevant responsible persons based on the time nodes set in the plan. Examination and assessment steps: Create an examination and assess the examination results according to the training plan. By analyzing the matching degree between the candidates' answers and the knowledge points associated with the test questions, identify the employees' knowledge weaknesses. When there is multiple examination data, obtain the historical score rate by statistically analyzing the historical scores of the employees' knowledge weaknesses, and generate data containing the employees' knowledge weaknesses, their corresponding historical score rates, and examination results, and store them in the database. Personnel status monitoring steps: Establish and maintain an electronic personnel information database to automatically identify personnel onboarding, job transfer, return to work, promotion, and certificate expiration status. When a status change is detected, automatically generate and push the corresponding training task chain that requires the completion of three levels of safety education: company, factory, and work team. Personalized training steps: Receive the training plan or training task chain, and create and execute training tasks based on the employee's knowledge weaknesses, historical score rate and test scores obtained from the database. In the training content, the explanation of corresponding knowledge points is dynamically strengthened in a targeted manner. During online training, image recognition technology is used to verify the identity of participants and generate training records. Needs analysis and feedback steps: Based on the exam scores and employee knowledge gap data obtained from the database, and combined with the historical score rate analysis, the trend of knowledge mastery is analyzed. Data clustering technology is used to identify common knowledge gaps, and individual training needs suggestions are generated accordingly. These individual needs suggestions are then summarized by department to form factory-level and company-level training needs plans. In developing the training plan for the following year, the aforementioned demand plan will be used as a core input to set the training content themes and determine the key areas of the examination, thereby achieving closed-loop management.

[0007] Furthermore, in the aforementioned examination and assessment steps: The exam creation supports two modes: manual exam paper generation and AI exam paper generation. The exam results are assessed using a combination of AI-automated grading and human grading, and the system supports the withdrawal and re-grading of already graded papers.

[0008] Furthermore, the AI-powered test paper generation is implemented using natural language processing technology, specifically including: The system analyzes the safety regulations documents of the non-ferrous heavy metal enterprises, automatically extracts key knowledge points, and generates a question bank. It receives user configuration instructions on question types and quantities, and automatically generates test papers from the question bank according to the instructions.

[0009] Furthermore, the AI-assisted automatic grading of papers, for subjective questions, is implemented using a semantic similarity algorithm, specifically including: Vectorize the text of the test taker's answers and the standard answers; Calculate the semantic similarity between vectors; The score is determined according to the preset scoring rules based on segmented mapping.

[0010] Furthermore, in the personnel status monitoring step, the generation and delivery of the training task chain specifically includes: When a record of job transfer, return to work, or promotion is detected in the electronic personnel information database, a training task chain for safety education at the company, factory, and work team levels that needs to be completed in sequence is generated, and this task chain is pushed to the relevant person in charge as one of the inputs to trigger personalized training.

[0011] Furthermore, in the personalized training steps, the dynamic reinforcement of the explanation of corresponding knowledge points specifically includes: Based on the data on employees' knowledge gaps, specific knowledge points are identified. If the historical score rate for a knowledge point is lower than a preset threshold, the training materials generated for that employee will be used to dynamically strengthen the explanation of that knowledge point by increasing the length of the explanation, supplementing typical cases, or reinforcing warning content.

[0012] Furthermore, in the personalized training step, the method for creating training content includes: Natural language processing technology was used to analyze the training materials and extract core knowledge points and logical structures. Based on the extracted content, structured training courseware containing titles, key knowledge points, teaching cases, and summaries is automatically generated.

[0013] Furthermore, in the needs analysis and feedback step, the formation of the training needs plan specifically involves: Individual training needs suggestions are compiled by department, and common weaknesses are identified through data clustering; Based on this, a factory-level special training plan and a company-level key training plan are generated, which will serve as the basis for formulating the training plan for the following year.

[0014] Another objective of this invention is to provide an AI adaptive training system for non-ferrous heavy metal safety standardization information, used to implement the aforementioned AI adaptive training method for non-ferrous heavy metal safety standardization information, comprising a processor, a memory, and a computer program stored in the memory, wherein when the program is executed by the processor, the following modules are logically coordinated: The training plan module is used to execute the steps of the training plan; The examination and assessment module, connected to the training plan module, is used to execute the examination and assessment steps and generate and store relevant data; The personalized training module is connected to the training plan module, the examination and assessment module, and the personnel status monitoring module. It is used to receive the training plan from the training plan module or the training task chain from the personnel status monitoring module, obtain the data from the examination and assessment module, and execute the personalized training steps. The personnel status monitoring module is connected to the training plan module and the personalized training module, and is used to execute the personnel status monitoring steps. The requirements analysis and feedback module, connected to the examination and assessment module and the training plan module, is used to execute the requirements analysis and feedback steps to form a closed-loop management system.

[0015] The AI ​​adaptive training system and method for non-ferrous heavy metal safety standardization information provided by this invention have the following significant advantages compared with existing technologies: Precise and Personalized: By collecting and analyzing multi-dimensional data such as employee learning behavior and performance evaluations, the system can accurately pinpoint each employee's knowledge gaps, enabling highly customized training content. This transforms training from the traditional "broad-based" approach to "precision-based," significantly improving training efficiency and effectiveness while enhancing employee learning motivation and focus.

[0016] Efficient Management: The electronic and automated management of the entire training process greatly reduces the administrative burden on managers. It supports online operation across the entire chain, from planning and notification to result entry and file generation, freeing managers from tedious tasks and allowing them to devote more energy to the design and optimization of the training system, thereby improving overall management efficiency.

[0017] Closed-loop adaptive: Possesses strong self-optimization capabilities. Through continuous needs analysis and performance feedback mechanisms, it can dynamically identify changing trends in the capabilities of enterprises and employees, and automatically adjust the focus and content of training accordingly, ensuring that the training system always evolves in sync with safety needs, maintaining continuous effectiveness and foresight.

[0018] Strong Compliance and Risk Control: The course content and assessment standards are closely aligned with the latest safety regulations in the non-ferrous and heavy metals industry. Through continuous and intensive training and assessment, the system effectively enhances employees' safety awareness and operational standardization, expands the depth and breadth of safety management coverage, reduces the probability of safety accidents from the source, and provides a guarantee for the company's stable production. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] This embodiment provides an AI adaptive training method for non-ferrous heavy metal safety standardization information, implemented based on an AI adaptive training system for non-ferrous heavy metal safety standardization information. The hardware foundation of the system can be a server cluster consisting of one or more servers, or virtual server resources based on cloud computing. The system includes a processor (CPU), memory (RAM and hard disk storage devices), and a computer program stored on the memory. The system may also include necessary input / output devices, network interfaces, etc., to facilitate data interaction with user terminals (such as computers and mobile phones).

[0022] The aforementioned system is preferably deployed on an enterprise intranet or private cloud environment to ensure the security of sensitive training data and personnel information. In specific deployments, the system's natural language processing engine, image recognition engine, and data clustering analysis tasks can utilize GPU computing resources for parallel computation acceleration to meet the real-time requirements of analyzing large amounts of text, images, and data.

[0023] The core of the system lies in achieving closed-loop training management through the collaboration of software modules. These modules logically include: a training plan module, an examination and assessment module, a personalized training module, a personnel status monitoring module, and a needs analysis and feedback module. The modules exchange data and transmit instructions through predefined RESTful API interfaces and JSON data formats. For example, when the personnel status monitoring module detects a status change, it pushes a structured task data package to a designated interface of the personalized training module. This data package contains key information such as employee ID, change type, trigger timestamp, and the required training task chain, ensuring the accuracy and timeliness of data flow. Before implementing the method, it is necessary to systematically collect and organize safety regulations and training materials from non-ferrous heavy metal enterprises, such as the "Safety Operating Procedures for Copper Smelting," "Emergency Plan for Heavy Metal Poisoning," "Electrolytic Cell Safety Operation Manual," accident case studies, and equipment operation manual PPTs. These documents are the foundation for subsequent AI natural language processing and analysis.

[0024] Subsequently, knowledge graph technology is used to transform these unstructured documents into a structured knowledge network. The specific process includes: first, extracting key knowledge points such as "prevention of burns from molten metal at high temperatures" and "emergency handling of electrolyte leaks" using entity recognition technology; then, clarifying the logical relationships (such as "prerequisites," "operational steps," and "risk consequences") and hierarchical levels (such as "emergency handling of electrolyte leaks" belonging to the category "safe operation in electrolysis workshops") between knowledge points using relation extraction technology; and finally, constructing a semantically related knowledge graph. This structured knowledge system will serve as the core data foundation for AI-generated test papers, training content, and requirements analysis, enabling machines to "understand" the internal connections of safety knowledge, rather than simply storing text.

[0025] This embodiment provides an AI-adaptive training method for safety standardization information in non-ferrous heavy metals. Based on the safety regulations and training materials of non-ferrous heavy metal enterprises, it constructs a data-driven closed-loop training management system through the following steps: training planning, examination and assessment, personnel status monitoring, personalized training, and needs analysis and feedback. Specifically, when formulating the training plan for the following year, the needs assessment plan is used as the core input to set the training content themes and determine the key areas of examination, thus achieving closed-loop management. The specific process of the method is as follows... Figure 1 As shown below, the implementation process of the method will be explained in detail.

[0026] I. Training Plan Steps: Based on the three-tier organizational structure of company, factory, and work team, and in accordance with the preset approval path rules corresponding to the three-tier structure, an electronic training plan is generated. The plan is reviewed and released through the corresponding multi-level electronic approval process, and reminder messages are sent to relevant responsible persons based on the time nodes set in the plan.

[0027] This step is executed by the training plan module. Based on the three-tier organizational structure of company, factory, and work team, and according to the preset approval path rules corresponding to the three-tier structure, an electronic training plan is generated.

[0028] Plan Creation and Approval: Taking "Annual Heavy Metal Protection Knowledge Training" as an example, the system will automatically match the approval path according to the plan level. Company-level plans require three levels of electronic approval: the head of the safety management department, the leader in charge of safety, and the company's main person in charge; factory-level plans require approval from the factory's safety manager and the factory's head. All approval processes are completed online, and approval comments and timestamps are recorded.

[0029] Plan Monitoring and Reminders: After a plan is published, the system's backend service continuously monitors the time nodes for each plan item. For example, it can automatically send in-system messages or email reminders to the plan manager one week and one day before the plan starts. This workflow-based automated approval and monitoring significantly improves the efficiency and standardization of plan management, ensuring the timely start of training activities.

[0030] II. Examination and Assessment Steps: Based on the training plan, create examinations and assess scores. By analyzing the matching degree between candidates' answers and the relevant knowledge points in the test questions, identify employees' knowledge weaknesses. When multiple examination data exist, calculate the historical score rate by statistically analyzing the historical scores of the employees' knowledge weaknesses, and generate data containing the employees' knowledge weaknesses, their corresponding historical score rates, and examination scores, which is then stored in the database. The examination creation process supports both manual and AI-generated test paper modes. This step is performed by the Examination and Assessment module and is a key step in generating core assessment data.

[0031] The AI-powered test paper generation is achieved through natural language processing (NLP) technology. This includes: parsing the safety regulations documents of the non-ferrous heavy metal enterprises, automatically extracting key knowledge points and generating a question bank; receiving user configuration instructions on question types and quantities, and automatically generating a test paper from the question bank based on these instructions. Specifically, it calls a natural language processing (NLP) engine, which first parses the safety regulations documents based on a pre-trained BERT model, automatically extracting key knowledge points (such as extracting "ventilation detection," "oxygen concentration standards," and "rescue principles" from "confined space operation procedures") and generating an initial question bank. Users (administrators) can further configure question types (multiple choice / true / false) and the number of questions, and the system then intelligently generates a comprehensive test paper from the question bank.

[0032] The manual test paper assembly refers to the method by which administrators or examiners select and freely combine questions from the question bank according to clear teaching objectives and assessment requirements to create test papers. This mode provides core examination staff with absolute control over the content and structure of the test papers and is suitable for scenarios that require precise control over the content of test questions, such as routine tests and targeted exercises.

[0033] The exam scoring employs a combination of AI-automated grading and human grading, and supports the withdrawal and re-grading of already graded papers. The human-machine collaborative grading process utilizes a layered processing strategy, combining the efficiency and accuracy of AI with the professional judgment of humans. The entire process begins with initial AI grading, where the system automatically scores objective questions to ensure rapid processing of basic question types. For subjective questions, a semantic similarity algorithm is used to analyze the text, converting the student's answer and the standard answer into semantic vectors and calculating their similarity. A preliminary score is then derived based on pre-defined segmentation mapping rules. Simultaneously, AI automatically identifies papers with low-confidence scores, abnormal answer patterns, or those borderline passing grades, providing a focus for subsequent human intervention.

[0034] Next comes the stage of intensive review by examiners. Based on the clusters of papers requiring special attention identified in the AI ​​pre-marking, examiners quickly confirm or correct the issues using the AI-provided scoring suggestions and anomaly annotations, and provide a brief explanation for the corrections. These manually confirmed or corrected scoring results will be stored in the database as the final reliable data.

[0035] For situations involving disputes or requiring higher-level judgment, an arbitration and model optimization mechanism is in place. If a candidate or grader disagrees with the scoring results, they can initiate an arbitration application. In this case, a higher-authority arbitrator (such as a subject matter expert group or the head of the grading team) will conduct the final review. More importantly, all manually corrected records and arbitration results will be fed back into the knowledge base of the AI ​​grading model as high-quality labeled data. This data will be used to continuously optimize the accuracy of semantic similarity calculations and the adaptability of the scoring rules, thus forming a self-improving closed-loop learning system.

[0036] To ensure the seriousness and traceability of the scoring results, the process also includes a result confirmation and locking step. Once the scoring arbitration process is complete, the person in charge of the assessment can make a final confirmation of the overall examination results. After confirmation, the system will write-protect and lock the scores for all questions to prevent any unauthorized modifications. Finally, the system automatically summarizes the scores for each question and generates each candidate's total score.

[0037] This process design, by clearly defining the responsibilities of both humans and machines and focusing human review on the aspects requiring the most professional judgment, not only significantly improves marking efficiency but also effectively ensures the accuracy, fairness, and reliability of the scoring results through a closed-loop learning mechanism and rigorous process management. The AI-powered automatic grading system uses a semantic similarity algorithm for subjective questions. This involves: vectorizing the candidate's answer and the standard answer text; calculating the semantic similarity between the vectors; and determining the score according to a pre-defined, segmented mapping-based scoring rule. After the exam, for objective questions, the system directly compares the score with the standard answer. For subjective questions (such as "briefly describe the evacuation procedures in the event of an arsine leak"), a semantic similarity algorithm is used for grading. Specifically, the Sentence-BERT model is used to convert the candidate's answer and the standard answer text into 384-dimensional semantic vectors, and cosine similarity is calculated. The system pre-defines segmented scoring rules; for example, a similarity greater than or equal to 0.8 is considered excellent (100% full marks), 0.6 to 0.8 is considered passing (60% marks), and below 0.6 is considered failing (no marks). This scoring rule can be configured in the system backend according to the importance of different questions. All score data, along with employee knowledge weaknesses identified by analyzing candidates' performance on various knowledge points and historical score rates calculated from statistical historical exam data, are structured and stored in a database. This data system provides a precise basis for subsequent personalized training.

[0038] To ensure the reversibility and traceability of exam score evaluation, a complete workflow for withdrawal and re-grading was designed. Triggering conditions and permissions: Active initiation: If an examiner discovers a misjudgment, they can withdraw the papers they have graded within their authority.

[0039] Candidate Appeal Triggered: Candidates submit a review application through the system with reasons. After the application is approved, the system will automatically reset the exam paper status to "pending grading" and notify the original grading teacher or arbitrator.

[0040] Administrator forced withdrawal: Administrators with advanced privileges can forcibly withdraw the marking results of any exam paper, which is usually used to handle major disputes or systemic issues.

[0041] III. Personnel Status Monitoring Steps: Establish and maintain an electronic personnel information database to automatically identify personnel onboarding, job transfers, return to work, promotions, and certificate expiration statuses. When a status change is detected, automatically generate and push a corresponding training task chain requiring sequential completion of company, factory, and work team level safety education. The generation and pushing of the training task chain specifically includes: when a job transfer, return to work, or promotion record is detected in the electronic personnel information database, generate a training task chain requiring sequential completion of company, factory, and work team level safety education, and push this task chain as one of the inputs to trigger personalized training to the relevant responsible person.

[0042] The personnel status monitoring module interfaces with the human resources system to monitor the electronic personnel information database in real time. When a change in an employee's status is detected (such as a transfer from "smelting worker" to "electrolytic refining worker"), the module will immediately and automatically generate a training task chain that includes safety education at the company, plant, and work team levels, and push it to the personalized training module and relevant personnel.

[0043] IV. Personalized Training Steps: Receive the training plan or training task chain, and create and execute training tasks based on the employee's knowledge weaknesses, historical score rate and exam results obtained from the database. In the training content, dynamically strengthen the explanation of corresponding knowledge points in a targeted manner. During online training, use image recognition technology to verify the identity of participants and generate training records.

[0044] Once the personalized training module receives a training plan or task chain, it initiates the creation and delivery of training content.

[0045] Courseware generation: Using NLP technology to analyze training materials (such as PPT and WORD documents), extract core knowledge points and logical structures, and automatically generate structured training courseware.

[0046] The dynamic reinforcement of the explanation of corresponding knowledge points specifically includes: identifying specific knowledge points based on the employee's knowledge weakness data; if the historical score rate of this knowledge point is lower than a preset threshold, then in the training courseware generated for the employee, the explanation of this knowledge point is dynamically reinforced by increasing the explanation length, supplementing typical cases, or strengthening warning content. The module will retrieve the employee's knowledge weaknesses and historical score rates from the database. For example, if the system identifies that employee A's historical score rate for the knowledge point "emergency eyewash station usage" is consistently lower than the preset threshold of 40%, then in the training courseware generated for him / her, the explanation length of this knowledge point will be automatically increased, and a relevant accident case video will be added for reinforcement.

[0047] The method for creating training content includes: using natural language processing technology to parse the training materials and extract core knowledge points and logical structures; and automatically generating structured training courseware containing titles, key knowledge points, teaching cases, and summaries based on the extracted content.

[0048] Authentication and Recording: During online training, the system utilizes the camera and image recognition technology based on MTCNN face detection and FaceNet feature comparison to perform seamless authentication during the training process, preventing proxy learning and testing. Upon completion of the training, the system automatically generates an electronic record containing the trainee's information, time, content, and evaluation results.

[0049] V. Needs Analysis and Feedback Steps: Based on the exam scores and employee knowledge gap data obtained from the database, and combined with the historical score rate analysis to analyze the changing trends in the mastery of knowledge points, data clustering technology is used to identify common knowledge gaps, and individual training needs suggestions are generated accordingly. These individual needs suggestions are then summarized by department to form factory-level and company-level training needs plans.

[0050] The formation of the training needs plan specifically involves: summarizing individual training needs suggestions by department, identifying common weaknesses through data clustering, generating factory-level special training plans and company-level key training plans based on these, and using them as the basis for formulating the training plan for the following year.

[0051] The requirements analysis and feedback module periodically (e.g., quarterly) extracts all employees' exam scores and knowledge gap data from the database, combining this with historical score rates to analyze trends in knowledge mastery. Subsequently, data clustering algorithms (such as K-Means clustering) are used to identify common knowledge gaps from the massive dataset. For example, cluster analysis might reveal that "employees in Workshop 2 generally have a weakness in the knowledge point of 'Identification of Occupational Hazards Caused by Heavy Metals'." Based on this, the system generates individual training needs suggestions, which are then aggregated by department, ultimately forming a factory-level specialized training plan and a company-level key training program.

[0052] These data-driven needs assessments serve as core inputs when developing the training plan for the following year. Training planners can clearly see which content needs strengthening and which exam focuses need adjustment, thus creating a more targeted annual training plan. This forms a complete, continuously self-optimizing closed-loop management system of "planning-execution-evaluation-feedback-plan optimization," ensuring that training efforts consistently respond to the actual needs of the organization and its employees.

[0053] This embodiment also provides an AI adaptive training system for non-ferrous heavy metal safety standardization information, used to implement the aforementioned AI adaptive training method for non-ferrous heavy metal safety standardization information, including a processor, a memory, and a computer program stored in the memory. When the program is executed by the processor, it logically implements the coordination of the following modules: The training plan module is used to execute the steps of the training plan; The examination and assessment module, connected to the training plan module, is used to execute the examination and assessment steps and generate and store relevant data; The personalized training module is connected to the training plan module, the examination and assessment module, and the personnel status monitoring module. It is used to receive the training plan from the training plan module or the training task chain from the personnel status monitoring module, obtain the data from the examination and assessment module, and execute the personalized training steps. The personnel status monitoring module is connected to the training plan module and the personalized training module, and is used to execute the personnel status monitoring steps. The requirements analysis and feedback module, connected to the examination and assessment module and the training plan module, is used to execute the requirements analysis and feedback steps to form a closed-loop management system.

[0054] In summary, this invention, through the synergistic effect of the above steps, constructs a data-driven closed-loop training management system. This system enables personalized dynamic adjustment of training content and continuous optimization of training plans, thereby improving the relevance of training and management efficiency. Its beneficial effects include at least the following aspects: Precise and personalized: Through multi-dimensional data collection and analysis, the training content is precisely matched with the employees' knowledge gaps, transforming the traditional "broad-based" training into "precision drip irrigation," significantly improving training efficiency.

[0055] Efficient management: The entire process is electronic and automated, which greatly reduces the administrative burden on training managers, allowing them to focus more on the design and optimization of the training system.

[0056] Closed-loop adaptive: The system has the ability to self-optimize. Through the demand analysis and feedback mechanism, it can ensure that the training content and focus are dynamically adjusted as the company's safety needs and employees' capabilities change, thus maintaining the continuous effectiveness of the training system.

[0057] Strong compliance and risk control: Closely aligning with the safety standards of the non-ferrous and heavy metals industry, and through continuous reinforcement and assessment, effectively enhance employees' safety awareness and operational standardization, thereby reducing the risk of safety accidents from the source.

[0058] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0059] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0060] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0061] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0062] Contents not described in detail in this specification are prior art known to those skilled in the art. It is hereby indicated that the above description is intended to help those skilled in the art understand this invention, but does not limit the scope of protection of this invention. Any equivalent substitutions, modifications, improvements, or simplifications of the above descriptions that do not depart from the essential content of this invention fall within the scope of protection of this invention.

Claims

1. An AI adaptive training method for safety standardization information of non-ferrous heavy metals, characterized in that, The method, based on the safety regulations and training materials of non-ferrous heavy metal enterprises, constructs a data-driven closed-loop training management system through the following steps: Training plan steps: Based on the three-level organizational structure of company, factory and work team, and according to the preset approval path rules corresponding to the three-level structure, an electronic training plan is generated. The plan is reviewed and released through the corresponding multi-level electronic approval process, and reminder messages are sent to relevant responsible persons based on the time nodes set in the plan. Examination and assessment steps: Create an examination and assess the examination results according to the training plan. By analyzing the matching degree between the candidates' answers and the knowledge points associated with the test questions, identify the employees' knowledge weaknesses. When there is multiple examination data, obtain the historical score rate by statistically analyzing the historical scores of the employees' knowledge weaknesses, and generate data containing the employees' knowledge weaknesses, their corresponding historical score rates, and examination results, and store them in the database. Personnel status monitoring steps: Establish and maintain an electronic personnel information database to automatically identify personnel onboarding, job transfer, return to work, promotion, and certificate expiration status. When a status change is detected, automatically generate and push the corresponding training task chain that requires the completion of three levels of safety education: company, factory, and work team. Personalized training steps: Receive the training plan or training task chain, and create and execute training tasks based on the employee's knowledge weaknesses, historical score rate and test scores obtained from the database. In the training content, the explanation of corresponding knowledge points is dynamically strengthened in a targeted manner. During online training, image recognition technology is used to verify the identity of participants and generate training records. Needs analysis and feedback steps: Based on the exam scores and employee knowledge gap data obtained from the database, and combined with the historical score rate analysis, the trend of knowledge mastery is analyzed. Data clustering technology is used to identify common knowledge gaps, and individual training needs suggestions are generated accordingly. These individual needs suggestions are then summarized by department to form factory-level and company-level training needs plans. In developing the training plan for the following year, the aforementioned demand plan will be used as a core input to set the training content themes and determine the key areas of the examination, thereby achieving closed-loop management.

2. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 1, characterized in that, In the examination and assessment steps described: The exam creation supports two modes: manual exam paper generation and AI exam paper generation. The exam results are assessed using a combination of AI-automated grading and human grading, and the system supports the withdrawal and re-grading of already graded papers.

3. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 2, characterized in that, The AI-powered test paper generation is achieved through natural language processing technology, specifically including: The system analyzes the safety regulations documents of the non-ferrous heavy metal enterprises, automatically extracts key knowledge points, and generates a question bank. It receives user configuration instructions on question types and quantities, and automatically generates test papers from the question bank according to the instructions.

4. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 2, characterized in that, The AI-assisted automatic grading, for subjective questions, is implemented using a semantic similarity algorithm, specifically including: Vectorize the text of the test taker's answers and the standard answers; Calculate the semantic similarity between vectors; The score is determined according to the preset scoring rules based on segmented mapping.

5. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 1, characterized in that, In the personnel status monitoring step, the generation and push of the training task chain specifically includes: When a record of job transfer, return to work, or promotion is detected in the electronic personnel information database, a training task chain for safety education at the company, factory, and work team levels that needs to be completed in sequence is generated, and this task chain is pushed to the relevant person in charge as one of the inputs to trigger personalized training.

6. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 1, characterized in that, In the personalized training process, the dynamic reinforcement of the explanation of corresponding knowledge points specifically includes: Based on the data on employees' knowledge gaps, specific knowledge points are identified. If the historical score rate for a knowledge point is lower than a preset threshold, the training materials generated for that employee will be used to dynamically strengthen the explanation of that knowledge point by increasing the length of the explanation, supplementing typical cases, or reinforcing warning content.

7. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 1, characterized in that, In the personalized training process, the methods for creating training content include: Natural language processing technology was used to analyze the training materials and extract core knowledge points and logical structures. Based on the extracted content, structured training courseware containing titles, key knowledge points, teaching cases, and summaries is automatically generated.

8. The AI ​​adaptive training method for non-ferrous heavy metal safety standardization information according to claim 1, characterized in that, In the needs analysis and feedback step, the formation of the training needs plan specifically involves: Individual training needs suggestions are compiled by department, and common weaknesses are identified through data clustering; Based on this, a factory-level special training plan and a company-level key training plan are generated, which will serve as the basis for formulating the training plan for the following year.

9. An AI adaptive training system for non-ferrous heavy metal safety standardization information, used to implement the AI ​​adaptive training method for non-ferrous heavy metal safety standardization information as described in any one of claims 1-8, characterized in that, It includes a processor, a memory, and a computer program stored in the memory, which, when executed by the processor, logically implements the coordination of the following modules: The training plan module is used to execute the steps of the training plan; The examination and assessment module, connected to the training plan module, is used to execute the examination and assessment steps and generate and store relevant data; The personalized training module is connected to the training plan module, the examination and assessment module, and the personnel status monitoring module. It is used to receive the training plan from the training plan module or the training task chain from the personnel status monitoring module, obtain the data from the examination and assessment module, and execute the personalized training steps. The personnel status monitoring module is connected to the training plan module and the personalized training module, and is used to execute the personnel status monitoring steps. The requirements analysis and feedback module, connected to the examination and assessment module and the training plan module, is used to execute the requirements analysis and feedback steps to form a closed-loop management system.

Images