A safety training model, method, system, and computer medium based on a capability assessment dynamic weight optimization model.

Abstract

This invention provides a safety training model, method, system, and computer medium based on a dynamic weight optimization model for competency assessment. Specifically, it constructs a three-dimensional graph model of competency, knowledge, and course, establishes a three-dimensional mapping relationship to determine initial weights, and then dynamically integrates, optimizes, and triggers subjective and objective weights based on trainee behavior data. A personalized course recommendation sequence is generated through a reinforcement learning algorithm. The state space is composed of competency assessment vectors and knowledge mastery vectors, and the course resource set is the action space. The recommendation strategy is constructed using short-term learning efficiency and long-term job assessment pass rate as reward functions. The invention also provides a computer medium based on the above scheme. This invention can provide chemical enterprise employees with targeted and accurate safety competency assessment methods and training courses.

Description

Technical Field

[0001] This invention belongs to the technical field of chemical safety training model systems and methods, and particularly relates to a safety training model, method and system, and computer medium based on a dynamic weight optimization model for capability assessment. Background Technology

[0002] In high-risk industries such as chemicals, safe production is of paramount importance, and the cultivation of professional personnel and the improvement of safety capabilities are key to ensuring safe production. The chemical industry uses a wide variety of hazardous chemicals and involves complex processes, making accidents extremely dangerous. With the continuous advancement of technology in these industries, production equipment is becoming increasingly complex, placing ever higher demands on the safety skills of employees. Safety production training is a fundamental task in the production process of chemical enterprises. As my country continues to promote the optimization and upgrading of its industrial structure, the level of mechanization, automation, digitalization, and intelligence in the chemical industry will continue to improve. Production equipment will become more complex, processes more dangerous, and safety hazards more hidden and complex. There is an urgent need for highly qualified personnel with specialized skills, strong business acumen, and excellent learning abilities to enter the chemical industry and promote its stable and orderly development.

[0003] While numerous research findings exist regarding the establishment and evaluation of safety competence models, significant shortcomings remain. In model building, existing studies often target specific industries or positions, lacking versatility and making direct application to other sectors difficult. Regarding safety competence assessment, methods are often complex and imprecise, failing to effectively meet enterprises' needs for rapid and accurate evaluation of employee safety skills. Some scholars, based on accident analysis, have identified accident-prone characteristics in current mining operations. They have utilized text analysis, interviews, questionnaires, and statistical analysis, employing a combination of qualitative and quantitative methods to construct a behavioral-characteristic safety competence model. Other scholars, considering the characteristics of work in power companies, have developed a hypothetical model of employee safety competence using the iceberg model theory combined with principal component analysis, providing a theoretical basis for personnel selection. Still others have addressed the issue of the proportion of human factors in accidents under the trend of power plant automation, analyzing the safety behavior of key power plant operators and constructing a personnel cognition model and an evaluation model based on multi-level matter-element analysis. In the same year, some scholars analyzed the relationships between factors influencing safety competence and, based on questionnaire survey results, proposed a model based on… DEMATEL The method of modeling safety capabilities of workers was used to identify the key influencing factors of safety capabilities of workshop workers.

[0004] Regarding safety capability assessment, some scholars have proposed basing it on the concept and connotation of safety capability of operators in machining workshops, in order to avoid... AHPTo address the shortcomings of the current method in fuzzy judgment, a hierarchical safety capability model is established based on the fuzzy hierarchical weighting method. Some scholars have proposed a three-factor model of construction workers' safety capability based on competency theory and safety theory, combined with literature review, questionnaire surveys, and behavioral event interviews, and established a safety capability evaluation system using structural equation modeling. Other scholars have proposed using questionnaires to survey personnel and apply [methods] to determine the relationship between safety atmosphere and subway construction worker safety. SPSS and AMOS The software obtained a safety relationship model after verifying and correcting the survey data. In the same year, some scholars proposed to establish a safety capability assessment index system with a tree-like hierarchical structure, and concluded that the safety capability of air traffic control units is formed by the interaction between five factors in the production process: personnel, equipment, environment, management, and transportation.

[0005] Currently, the development of talent in the chemical industry still faces shortcomings in terms of quantity, structure, quality, specialization, training, and support, failing to meet the current needs of the industry. Several problems and difficulties remain, such as: 1) a relative shortage of chemical professionals and a need to improve their professional skills; 2) insufficient emphasis on talent development in the chemical industry, resulting in training that lacks relevance and practicality, failing to meet the safety production needs of enterprises; and unsystematic training courses with a single, unsystematic approach, lacking a unified, scientific, and diversified training method; and 3) a lack of effective assessment of the safety skills of chemical practitioners, making it impossible to establish an effective assessment and evaluation system.

[0006] Therefore, in response to the existing practical problems, providing chemical enterprise employees with targeted and accurate safety capability assessment methods and training courses, and training high-quality chemical talents for chemical enterprises, is of great significance to the level of chemical safety production. Summary of the Invention

[0007] Technical solution: In order to solve the above-mentioned technical problems, the present invention provides a safety training model based on a capability assessment dynamic weight optimization model, the model including a capability assessment module, a dynamic weight calculation module, a reinforcement learning recommendation module and a real-time feedback module;

[0008] The competency assessment module is configured to extract three-dimensional indicators of basic competency, professional competency, and management competency required for the employee's position based on the job competency model, and generate a personalized competency assessment vector by combining the employee's contextual characteristics. The competency assessment vector includes the label level and initial weight of each competency dimension.

[0009] The dynamic weight calculation module is configured to construct a three-dimensional relationship between "ability-knowledge-course" and perform weight optimization and adjustment. The weight optimization includes establishing an ability-knowledge mapping matrix, a knowledge-course mapping relationship, and a fusion of subjective and objective weights.

[0010] The reinforcement learning recommendation module uses the ability assessment vector and the knowledge mastery vector to form the state space. The course resources are considered as an action space, with short-term learning efficiency (the ratio of learning duration to mastery) and long-term job performance evaluation pass rate serving as the reward function. Based on the reinforcement learning framework and policy gradient algorithm, a recommendation strategy is constructed, and the course recommendation order is given by dynamically optimizing the Q-value iteration.

[0011] The real-time feedback module is configured to collect student behavior data and contextual features in real time and trigger a set mechanism.

[0012] The model uses a three-dimensional graph to structurally link job competencies, professional knowledge systems, and training course resources, forming a dynamically evolving competency development path.

[0013] As an improvement, the competency assessment module selects structural equation modeling analysis for index selection. It uses structural equation modeling to verify the safety competency of practitioners, indirectly reflecting previously unobservable latent variables, discovering the relationship between latent variables, reflecting the relationship between manifest and latent variables, setting evaluation indicators, and establishing a model of the relationship between variables.

[0014] The structural equation model includes a measurement model and a structural model; specifically...

[0015] (1)

[0016] (2)

[0017] (3)

[0018] In equations (1)-(3), In the model, it is represented as the first... j One exogenous manifest variable, Represented as the first j One endogenous explicit variable Represented as the first i An exogenous latent variable, Represented as the first i , One endogenous latent variable, Represented as exogenous manifest variables In exogenous latent variables Factor loading coefficient matrix on Λy ij Indicates endogenous manifest variables In endogenous latent variables Factor loading coefficient matrix on Represented as exogenous manifest variables The error; Representing endogenous latent variables and The path coefficient matrix between them; Representing exogenous latent variables Endogenous latent variables The impact; Represents the residual term; Indicates endogenous potential The error.

[0019] As an improvement, the exogenous explicit variables include exam scores, training attendance rates, simulation operation scores, and fault handling speed; the endogenous explicit variables include emergency drill scores, accident handling scores, safety behavior self-assessment, accident incidence rate, and safety inspection scores; the endogenous latent variables include emergency response capabilities, safety attitudes, and safety performance; and the exogenous latent variables include safety knowledge and operational skills.

[0020] As an improvement, specific ways to configure the "ability-knowledge-curriculum" three-dimensional relationship include:

[0021] (1) Establish a capability-knowledge mapping matrix and determine the initial association weights between capability dimensions and knowledge units based on expert knowledge graphs and historical training data mining;

[0022] (2) Establish the knowledge-course mapping relationship, and determine the coverage weight of knowledge units and course resources by analyzing the course outline, marking the knowledge points of test questions, and back-inferring the learning effect;

[0023] (3) Integrate subjective and objective weights and set up a dynamic adjustment formula. Subjective weight Based on the Analytic Hierarchy Process AHP Alternatively, the Delphi method can be used to determine objective weights. Based on students' answer accuracy, learning time, and simulated operation scores, dynamic calculations are performed using a data association algorithm, with weighting adjustment factors. The decay factor decreases exponentially with increasing training data volume; , t For the number of iterations, This is the attenuation coefficient, which has a value greater than 0 and is used to control the weight adjustment factor. The rate at which the training data decays with increasing iterations is, in safety capability assessment, where λ can be dynamically adjusted based on the amount of accident case data. When historical accident data is limited (e.g., ...), the decay rate can be adjusted when there is little historical accident data (e.g., ...). t <10), Higher, preferred AHP Determine the weight of security knowledge; once enough data is accumulated (e.g.) t >50), Approaching 0, the model mainly calculates weights based on objective data such as trainee operation scores and accident handling records.

[0024] As an improvement, in the reinforcement learning recommendation module, the state space... ,in Indicates the first i Evaluation values ​​for each capability dimension Indicates the first j Mastery of each knowledge unit; reward function in , These are the weighting coefficients. These are the assessment values ​​for the ability dimensions before and after learning, respectively. For study time, The pass rate for job performance evaluation.

[0025] As an improvement, the contextual features include at least one of the following: years of service, job type, position type, position risk level, and historical training records; the reinforcement learning framework includes deep learning. Q network DQN Policy gradient algorithm PolicyGradient At least one; the behavioral data includes the accuracy rate of answering questions, learning time, interactive operation scores, and number of video replays.

[0026] As an improvement, the established mechanisms include:

[0027] (1) Update the importance of knowledge nodes by collecting real-time behavioral data on student operations, assessment scores, and knowledge mastery. Utilize this collected data to... PageRank The algorithm dynamically adjusts the node weights in the knowledge point network, updates the importance of knowledge nodes, establishes a safety capability assessment system for chemical industry practitioners, and establishes a safety training model based on the dynamic weight optimization model of capability assessment. It comprehensively considers factors such as employees' psychological qualities, business and technical capabilities, and safe operation skills, providing a basis for safety assessment, pre-training assessment, and post-training effectiveness verification.

[0028] (2) Abnormal knowledge node detection: When the standard deviation of the student's answer accuracy rate is >0.3, or the set knowledge points for detection are consecutive N When the error rate in answering questions exceeds the threshold, a weight adjustment and a reconstruction of the course resource association relationship are triggered.

[0029] (3) The weight adjustment factor is adaptive. The ratio of subjective and objective weight fusion is dynamically adjusted according to the data sample size. In the initial sample size, the expert experience weight is given priority, and the proportion of data-driven weight is gradually increased as the data accumulates.

[0030] As a specific embodiment of the present invention, a safety training method based on a capability assessment dynamic weight optimization model is also provided, the method comprising:

[0031] (1) Construct a three-dimensional graph model of competence-knowledge-course. In the three-dimensional graph model, the competence dimension is decomposed based on the job competency model and the weight is quantified. The knowledge dimension uses knowledge graph technology to construct a knowledge point network. The course dimension annotates the training resource library with metadata.

[0032] (2) Dynamically adjust subjective and objective weights based on student behavior data. Determine the initial weights by establishing a capability-knowledge mapping matrix and a knowledge-course mapping relationship, and adopt a subjective and objective weight fusion formula. Perform weight optimization, where The decay rate increases exponentially with the amount of training data.

[0033] (3) A personalized course recommendation sequence is generated by reinforcement learning algorithm. The state space is composed of ability assessment vector and knowledge mastery vector, the course resource set is the action space, and the recommendation strategy is constructed using short-term learning efficiency and long-term job assessment pass rate as reward functions.

[0034] As a specific embodiment of the present invention, the present invention provides a safety training system based on a capability assessment dynamic weight optimization model. This system operates based on the aforementioned safety training model and includes an initialization phase, an operation phase, and an optimization iteration phase, specifically:

[0035] (1) Initialization phase:

[0036] (1.1) Input personalized ability assessment vector into the ability assessment module This includes the level and initial weight of each competency dimension, and outputs a personalized competency assessment vector with the dimension label levels and initial weights generated by the job competency model, as well as the optimized "competency-knowledge-course" mapping weights. ;

[0037] (1.2) Exogenous and endogenous manifest variables were screened out using structural equation modeling;

[0038] (1.3) The ability assessment module passes the assessment values ​​of each ability dimension in the ability assessment vector to the reinforcement learning recommendation module. Including the initial mastery level of each knowledge unit （K 1 ,K 2 ,…,K n ） As a reinforcement learning state space The initial value;

[0039] (1.4) The dynamic weight optimization module sends the initial weights of the constructed three-dimensional relationship between “ability-knowledge-course” to the reinforcement learning recommendation module, including the initial weight data of the ability-knowledge mapping matrix and the knowledge-course mapping relationship;

[0040] (2) Operation phase:

[0041] (2.1) The reinforcement learning recommendation module calls the interface in real time through the dynamic weight optimization module to request the latest "ability-knowledge-course" three-dimensional relationship weight data for recommendation strategy calculation;

[0042] (2.2) The dynamic weight optimization module returns the basis through the reinforcement learning recommendation module. Calculated fusion of subjective and objective weights ,in , The attenuation coefficient;

[0043] (2.3) The reinforcement learning recommendation module pushes a personalized course recommendation sequence generated based on the reinforcement learning framework and policy gradient algorithm to the real-time feedback module. This sequence is then processed by the system. Q Value iteration and dynamic optimization;

[0044] (2.4) The real-time feedback module transmits student behavior data and features back to the reinforcement learning recommendation module in real time; the real-time feedback module transmits student behavior data and assessment data to the ability assessment module, triggering the update of the ability assessment vector and knowledge mastery vector; the real-time feedback module sends student operation behavior data, assessment score data, and knowledge mastery data to the dynamic weight optimization module, triggering the update of knowledge node importance, and adjusting the weight of knowledge point network nodes through the PageRank algorithm; when the abnormal knowledge node detection conditions are met, the standard deviation of the answer accuracy rate is >0.3 or the set detection knowledge point's consecutive N-time answer error rate exceeds the threshold, the weight adjustment and course resource association relationship reconstruction are triggered.

[0045] (3) Optimization and iteration stage:

[0046] (3.1) The updated capability assessment vector, including the latest assessment values ​​of each capability dimension and the knowledge mastery vector, is sent to the dynamic weight optimization module to trigger weight optimization and adjustment.

[0047] (3.2) The dynamic weight optimization module pushes the optimized three-dimensional correlation weight data of "ability-knowledge-course" to the reinforcement learning recommendation module to update the recommendation strategy;

[0048] (3.3) The ability assessment module passes the updated ability assessment vector and knowledge mastery vector to the reinforcement learning recommendation module to update the reinforcement learning state space S;

[0049] (3.4) Real-time feedback module: Provides assessment values ​​of students' abilities before and after learning. Study time and job performance evaluation pass rate P Used to calculate the reward function The reinforcement learning recommendation module is based on this. Q Value iteration optimization recommendation strategy;

[0050] (3.5) Dynamic weight optimization module: Synchronize the change information of the weight adjustment factor ∝, and the real-time feedback module adjusts the data collection frequency and the threshold parameters of the triggering mechanism accordingly; when it is detected that the student's ability has significantly improved after learning a certain type of course, the change in the ability assessment value exceeds the set threshold γ or there is no significant improvement or it is lower than the threshold. δ At the same time, the real-time feedback module sends course effectiveness evaluation data to assist the dynamic weight optimization module in deeply calibrating the relationship between "ability-knowledge-course".

[0051] As another specific embodiment of the present invention, a computer medium is also provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the security training method based on a capability assessment dynamic weight optimization model as described above.

[0052] Beneficial effects: The safety training model and method based on a capability assessment dynamic weight optimization model proposed in this invention have the following advantages compared with conventional methods:

[0053] (1) The model in this invention includes four modules. Through closed-loop information interaction of the ability assessment module, dynamic weight optimization module, reinforcement learning recommendation module, and real-time feedback module, it achieves dynamic adaptation of the entire process of "assessment-recommendation-learning-feedback-optimization". Compared with conventional static recommendation models, it can more accurately capture learners' ability shortcomings and knowledge gaps. Specifically, it is reflected in:

[0054] The real-time data interaction between the competency assessment module and the dynamic weight optimization module enables the three-dimensional relationship between "competency-knowledge-course" to be dynamically adjusted as the learner's competency changes, thus avoiding recommendation bias caused by the fixed weights in conventional models.

[0055] The high-frequency information flow between the reinforcement learning recommendation module and the real-time feedback module can quickly iterate the recommendation strategy based on the student's real-time learning behavior (such as the accuracy of answering questions, the number of times the video is watched, etc.), so that the course recommendation is upgraded from "batch push" to "personalized adaptation", and the short-term learning efficiency (time to mastery ratio) is improved by more than 30%.

[0056] The real-time feedback module transmits data bidirectionally to the competency assessment module and the dynamic weight optimization module, triggering updates to the importance of knowledge nodes and detection of abnormal knowledge nodes. This ensures that the model can respond promptly to learners' learning bottlenecks. Compared to the periodic updates of conventional models, the problem response speed is improved by 50%, and the pass rate for long-term job assessments is increased by more than 20%.

[0057] (2) The multi-module collaborative interaction mechanism of the present invention solves the pain points of "disconnect between assessment and recommendation" and "feedback lagging behind the learning process" in conventional training models, and transforms the ability development path from "preset fixed route" to "dynamic evolution trajectory", which significantly improves the utilization efficiency of training resources and the matching accuracy of job competency.

[0058] (3) The "ability-knowledge-course" three-dimensional graph model constructed in this invention is based on the ability dimension, which is decomposed and weighted according to the job competency model. The knowledge dimension uses knowledge graph technology to construct a knowledge point network. The course dimension uses metadata annotation for the training resource library.

[0059] (4) The present invention annotates multi-dimensional metadata for courses, and realizes quantitative association with knowledge and ability through a "knowledge-course mapping matrix". In existing systems, ability, knowledge and courses are mostly unidirectionally associated with "ability → course", and the association relationship is fixed. The present invention forms a dynamic closed loop among the three: ability shortcomings drive knowledge weight adjustment, knowledge blind spots trigger course recommendations, course learning effects feed back into ability assessment, and weights are adjusted according to the amount of data through W. final The formula is automatically optimized. Furthermore, this invention offers advantages such as improved accuracy, strong dynamic adaptability, optimized resource utilization, and enhanced interpretability. Attached Figure Description

[0060] Figure 1 This is the dynamic weight optimization logic diagram of the present invention.

[0061] Figure 2 This is a flowchart of the three-dimensional map construction process of the present invention.

[0062] Figure 3 This is a partial screenshot of the training curriculum system of Embodiment 2 of the present invention.

[0063] Figure 4 Example 1 is a system architecture diagram of the present invention.

[0064] Figure 5 Example 2 is a system architecture diagram of the present invention. Detailed Implementation

[0065] The technical solutions in the embodiments of the present invention will be clearly and completely described below, so that those skilled in the art can better understand the advantages and features of the present invention, thereby making a clearer definition of the scope of protection of the present invention. The embodiments described in this invention are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0066] This invention provides a safety training model, method, and system for a dynamic weight optimization model for capability assessment, as well as a computer medium, which will be described and introduced in detail below.

[0067] (i) A safety training model based on a dynamic weight optimization model for capability assessment is provided.

[0068] This model of the present invention is applied to a safety training system for chemical enterprises. The model includes a capability assessment module, a dynamic weight calculation module, a reinforcement learning recommendation module, and a real-time feedback module.

[0069] (1.1) Competency Assessment Module

[0070] The specific implementation method used in the capability assessment module of this invention includes utilizing... DACUM (Developing A) Curriculum) By conducting job task analysis, the professional competence requirements for positions in chemical enterprises are described in a specific and clear manner, providing necessary and accurate basic information for the development of training curriculum systems and essential knowledge and skills information for the construction of professional curriculum systems.

[0071] Furthermore, this module of the present invention uses structural equation modeling to verify the safety capabilities of practitioners, indirectly reflecting previously unobservable latent variables, discovering the relationships between latent variables, reflecting the relationship between manifest and latent variables, setting evaluation indicators, and establishing a relationship model between variables.

[0072] The structural equation model of this invention comprises two parts: a measurement model and a structural model. The measurement equations and structural equations corresponding to the two models are as follows:

[0073] (1)

[0074] in, The exogenous manifest variables are represented in the model as exam scores, training attendance rates, simulation operation scores, and fault handling speed. This indicates that the exogenous latent variable is represented as security knowledge in the model. ), Operational skills ( )wait. Indicates exogenous manifest variables In exogenous latent variables Factor loading coefficient matrix on, for example Indicates the score of the safety exam ( Safety knowledge The degree of contribution; Indicates exogenous manifest variables Error, for example This indicates the portion of the safety exam score that was not explained by safety knowledge.

[0075] (2)

[0076] The variables represent endogenous explicit variables, which are expressed in the model as emergency drill scores, accident handling scores, safety behavior self-assessments, accident incidence rates, and safety inspection scores. This represents an endogenous latent variable, which is expressed as emergency response capability in the model. Safety attitude () ), safety performance ( )wait; Indicates endogenous manifest variables Y In endogenous latent variables η Factor loading coefficient matrix on, for example Indicates the score of the emergency drill ( ) on emergency response capabilities ( The representativeness of ); Indicates endogenous manifest variables Error, for example ε 1 This indicates the portion of the emergency drill score that was not explained by emergency response capabilities.

[0077] Structural models reflect the relationships between latent variables, and path diagrams can represent structural relationships. The mathematical expression of a structural model is as follows:

[0078] (3)

[0079] The path coefficient matrix representing the relationships between endogenous latent variables, for example Indicates emergency response capability ( ) on safety performance ( The direct impact of ) This indicates the effect of exogenous latent variables on endogenous latent variables, for example... Indicates safety knowledge ( ) on emergency response capabilities ( The direct impact of ) This represents the residual term, reflecting the portion of η that was not explained, for example... The emergency response capability residuals explained by the model.

[0080] (1.2) Dynamic weight calculation module

[0081] See Figure 1 The diagram shown is a dynamic weight optimization logic diagram of the dynamic weight calculation module of the present invention. It is configured to construct a three-dimensional relationship of "ability-knowledge-course" and perform weight optimization and adjustment. The weight optimization includes establishing an ability-knowledge mapping matrix, a knowledge-course mapping relationship, and determining, integrating, and triggering subjective and objective weights.

[0082] Example 1:

[0083] 1.2.a Constructing a three-dimensional relationship between "ability, knowledge, and curriculum"

[0084] One hundred reactor operators from a chemical company were selected as a sample. To cultivate their "safe operation ability", a three-dimensional relationship of "ability-knowledge-course" was constructed, and the dynamic weight changes were tracked during the three-month training period.

[0085] Table 1. Schematic diagram of the three-dimensional relationship between "ability-knowledge-curriculum"

[0086] Dimension elements Capability Dimension Safe operation capability (C1), emergency response capability (C2), equipment maintenance capability (C3) Knowledge Unit Reactor pressure control specifications (K1), emergency procedures for hazardous chemical leaks (K2), safety valve calibration standards (K3), equipment inspection specifications (K4) Course Resources Course A (Safety Operation Procedures for Reactors), Course B (Emergency Drills for Hazardous Chemical Leaks), Course C (Special Equipment Calibration Technology)

[0087] 1.2.b Capability-Knowledge Mapping Matrix Construction and Weight Optimization

[0088] (1) Initial weight determination (expert subjective weight) W expert )

[0089] Five senior engineers and security experts (with over 10 years of experience) were invited to rate the correlation between "ability" and "knowledge" (1-5 points, with 5 points indicating a strong correlation). The average score was then normalized. W expert .

[0090] Table 2 Data after initial weight determination

[0091] Capability Dimension K1 Reactor Pressure Control K2 Hazardous Chemical Leakage Emergency Response K3 safety valve calibration Safe operation capability (C1) 0.8 0.3 0.7 Emergency response capability (C2) 0.2 0.9 0.4

[0092] (2) Extract training data for this position over the past year (80 trainees), calculate the Pearson correlation coefficient between “improvement in knowledge unit mastery” and “improvement in ability assessment value”, and obtain W_data after normalization.

[0093] Table 3 shows the normalized W_data data.

[0094] Capability Dimension K1 Reactor Pressure Control K2 Hazardous Chemical Leakage Emergency Response K3 safety valve calibration Safe operation capability (C1) 0.75 (correlation coefficient 0.72) 0.25 (correlation coefficient 0.21) 0.65 (correlation coefficient 0.63) Emergency response capability (C2) 0.15 (correlation coefficient 0.13) 0.85 (correlation coefficient 0.81) 0.35 (correlation coefficient 0.32)

[0095] (3) Integration of subjective and objective weights

[0096] Initial stage (t=5, 5 iterations, small amount of data).

[0097] If we take λ=0.1 (attenuation coefficient), then ∝=e^(-0.1×5)=0.606, which is the weight after fusion.

[0098] Table 4 Weight data after initial fusion stage

[0099] Capability Dimension K1 Reactor Pressure Control K2 Hazardous Chemical Leakage Emergency Response K3 safety valve calibration Safe operation capability (C1) 0.606×0.8+0.394×0.75≈0.78 0.606×0.3+0.394×0.25≈0.28 0.606×0.7+0.394×0.65≈0.68 Emergency response capability (C2) 0.606×0.2+0.394×0.15≈0.18 0.606×0.9+0.394×0.85≈0.88 0.606×0.4+0.394×0.35≈0.38

[0100] Data accumulation phase (t=30, 30 iterations, sufficient data).

[0101] The weights after fusion (are more dependent on objective data).

[0102] Table 5 Weighted data after fusion during the data accumulation phase

[0103] Capability Dimension K1 Reactor Pressure Control K2 Hazardous Chemical Leakage Emergency Response K3 safety valve calibration Safe operation capability (C1) 0.05×0.8+0.95×0.75≈0.75 0.05×0.3+0.95×0.25≈0.25 0.05×0.7+0.95×0.65≈0.65

[0104] 1.2.c Knowledge-Course Mapping Relationship Construction and Weight Optimization

[0105] (1) Initial coverage weights (expert evaluation)

[0106] Table 6. Expert assessment of course coverage of knowledge units (0-1)

[0107] Knowledge Unit Course A (Operating Procedures) Course B (Emergency Drill) Course C (Verification Techniques) K1 Reactor Pressure Control 0.9 (Core Coverage) 0.2 (minor involvement) 0.3 (related mention) K2 Hazardous Chemical Leakage Emergency Response 0.3 (related mention) 0.9 (Core Coverage) 0.1 (Not involved) K3 safety valve calibration 0.2 (minor involvement) 0.1 (Not involved) 0.9 (Core Coverage)

[0108] (2) Dynamic adjustment (based on learning effect back-inference)

[0109] The system collects student data through a real-time feedback module: If a student's mastery of the K1 knowledge unit improves by 70% after learning Course A (the historical average is 50%), then the coverage weight of Course A for K1 is increased to 0.95; If a student answers questions incorrectly in the K2 knowledge unit three times in a row (error rate 70% > threshold 50%), the coverage weight of Course B for K2 is increased from 0.9 to 0.98, and the recommendation priority of Course B is increased.

[0110] 1.2.d Triggering Mechanism and Optimization Effects

[0111] Abnormal knowledge node detection: In the second month of training, it was found that the standard deviation of the correct answer rate for K1 (emergency procedure for hazardous chemical leaks) was 0.35 > 0.3, triggering the dynamic weight calculation part of the real-time feedback module.

[0112] Adjust the mapping weight between K and emergency response capability (C2) (from 0.88 to 0.92).

[0113] Reconstruct the relationship between course B and course K, and increase the class hours of course B (from 20% to 30%).

[0114] Results Verification: Three months later, the trainees in this position...

[0115] The average C1 safety operation capability assessment score improved by 40% (compared to 25% for routine training).

[0116] The average assessment value of emergency response capability (C2) increased by 55% (compared to 30% for regular training).

[0117] The pass rate for job performance evaluations increased from 65% to 88%.

[0118] See Figure 2 The diagram shown is a flowchart of the construction process of the three-dimensional map model of ability-knowledge-course in this invention. The three dimensions include ability dimension, knowledge dimension and course dimension. Furthermore, the ability dimension is based on the job competency model and is decomposed into basic ability, professional ability and management ability, which are quantified by weight.

[0119] Furthermore, the knowledge dimension utilizes knowledge graph technology to construct a network of knowledge points that map to capabilities, clarifying knowledge units such as "fire prevention and explosion protection principles" and "hazardous chemical storage specifications," as well as knowledge relationships, such as "fire prevention and explosion protection principles" being prerequisite knowledge for "gas properties."

[0120] Furthermore, the course dimension involves metadata annotation of the training resource library, covering attributes such as course duration, difficulty, and format, as well as the coverage of matching knowledge points and competency objectives. This includes designing a three-dimensional graph model framework, defining competency dimensions, decomposing basic competencies, professional competencies, and management competencies based on the job competency model, and quantifying them through weights; defining the knowledge dimension, constructing a knowledge point network that maps to competencies, and structuring it using knowledge graph technology; and defining the course dimension by annotating the training resource library with metadata.

[0121] exist Figure 2 In this invention, a three-dimensional competency-knowledge-course model is constructed through specific implementation methods. Specifically, the course dimension is established through a professional curriculum system and a general curriculum system. The professional curriculum system is typically generated through job task analysis and needs to be strongly correlated with employees' work scenarios; employees should learn what they are doing. The general curriculum system focuses on core competency analysis for system planning.

[0122] Based on the characteristics of job competencies, the specific content of knowledge items for each job is summarized, and the behavioral descriptions of each job competency item are extracted. Different levels of behavioral characteristics are described, realizing the transformation of job competency from abstract concepts to specific content. Furthermore, the specific content of knowledge and competencies for each job is analyzed, refined, integrated, and summarized to form training courses. Course outlines are researched and developed, and a training course library for chemical industry practitioners is constructed.

[0123] Based on the training course library, and taking job competency as the basis, training objectives as the foundation, and modular course management as the approach, the courses are stratified and classified. According to the characteristics and needs of employees at different development stages, the courses are reasonably distributed and combined, and the logical relationships between the courses are sorted out. This realizes the transformation from a course library to a course system, and ultimately establishes a multi-level, multi-scenario safety skills training course system of "job-competency-course".

[0124] Example 2:

[0125] See Figure 3 The following is a partial screenshot of the specific implementation method, where (1) the letters A, B, and C represent the level of mastery: A means knowing; B means operating independently; C means guiding others; (2) the numbers represent the period: 1 means the period is 1 year; 2 means the period is 2 years; 3 means the period is 3 years; (3) the training method: M1 refers to classroom teaching; M2 refers to classroom teaching + examination; M3 refers to meetings or self-study; M4 refers to practical exercises.

[0126] (1.3) Reinforcement learning recommendation module and real-time feedback module

[0127] In this invention, the reinforcement learning recommendation module uses an ability assessment vector and a knowledge mastery vector to form the state space, and the course resource set as the action space. Short-term learning efficiency (the ratio of learning duration to knowledge mastery) and long-term job performance pass rate are used as reward functions. A recommendation strategy is constructed based on a reinforcement learning framework and a policy gradient algorithm. Q The value iterative dynamic optimization course provides a recommended order;

[0128] The state space ,in Indicates the first i Evaluation values ​​for each capability dimension Indicates the first j Mastery of each knowledge unit;

[0129] reward function , These are the weighting coefficients. These are the assessment values ​​for the ability dimensions before and after learning, respectively. For study time, The pass rate for job performance evaluation.

[0130] Example 3:

[0131] Define a state space S (capability + knowledge), and define capability dimensions (3): C1 = Safe operation capability (0-10 points), C2 = Emergency response capability (0-10 points), C3 = Equipment maintenance capability (0-10 points). Define knowledge units (4): K1 = Reactor pressure control (mastery level 0-1), K2 = Hazardous chemical leak emergency response (0-1), K3 = Safety valve calibration (0-1), K4 = Equipment inspection procedures (0-1).

[0132] Furthermore, we obtain the initial state S0:

[0133] S0=[C1=5.5, C2=4.2, C3=6.8; K1=0.5, K2=0.3, K3=0.6, K4=0.7].

[0134] Furthermore, Action Space A (Course Resources):

[0135] Course A: Practical Operation of Reactor Pressure Control (related to C1 and K1);

[0136] Course B: Emergency Drill for Hazardous Chemical Spills (related to C2 and K2);

[0137] Course C: Safety Valve Calibration Standard (related to C3 and K3).

[0138] Then, the reward function R (ω1=0.5, ω2=0.5):

[0139]

[0140] in Δt Study time (hours); P Job performance evaluation pass rate (0-1).

[0141] The core function of the real-time feedback section is to collect data: learning time, answer accuracy, changes in knowledge mastery, and assessment results.

[0142] Triggering adjustment: When the mastery improvement rate of a certain knowledge unit is less than 30% or the error rate of answering questions is greater than 50%, a real-time warning will be pushed to the reinforcement learning module.

[0143] The following section will introduce and explain the above technical content through a detailed and complete process.

[0144] 1. Initial Recommendation (Based on State S0): The reinforcement learning module calculates the initial Q-value (Q(S0,A)), and the policy gradient algorithm selects the optimal course.

[0145] Q(S0,A)=2.8, Q(S0,B)=3.5, Q(S0,C)=2.2 → Recommended course B (emergency drills, addressing weaknesses in C2 and K2).

[0146] 2. Learning Process and Real-Time Feedback: The real-time feedback module collects data throughout the learning process of course B.

[0147] Learning time Δt = 2 hours, answer data: K2 related question accuracy = 60% (below the threshold of 70%), triggering an "abnormal knowledge node" warning. Note the changes in mastery: K2 increased from 0.3 to 0.5 (67% improvement), C2 increased from 4.2 to 5.0 (0.8 improvement).

[0148] 3. Reward Calculation and Q-Value Update

[0149] Short-term rewards (based on feedback data):

[0150]

[0151] Long-term reward (assessed after 1 month): The student passes the assessment (P=1) → 0.5×1=0.5.

[0152] The total reward R1 = 0.2 + 0.5 = 0.7.

[0153] Q-value iteration (Q-learning formula):

[0154] Original Q(S0,B)=3.5, updated Q(S0,B)=3.5+0.1×(0.7+0.9×maxQ(S1,A')-3.5) (α=0.1, γ=0.9).

[0155] (Note: S1 is the state after learning, and maxQ(S1,A') is the optimal course Q value under S1, assumed to be 3.2)

[0156] After the update, Q(S0,B) = 3.5 + 0.1 × (0.7 + 2.88 - 3.5) = 3.5 + 0.008 = 3.508.

[0157] 4. Status Updates and Secondary Recommendations (Driven by Real-Time Feedback)

[0158] The real-time feedback module pushes the message "K2 answer accuracy 60% < 70%" to the ability assessment module, triggering a status update.

[0159] New state S1 = [C1 = 5.5, C2 = 5.0, C3 = 6.8; K1 = 0.5, K2 = 0.5, K3 = 0.6, K4 = 0.7]

[0160] The reinforcement learning module recalculates the recommendation policy based on S1 and the updated Q-value.

[0161] Q(S1,A)=2.9, Q(S1,B)=3.6 (Q value increases because K2 still needs reinforcement), Q(S1,C)=2.3 → Course B is recommended again.

[0162] 5. Secondary learning and feedback adjustments

[0163] The student studied Course B again (the real-time feedback module adjusted the course content: added K2 practical cases), and the data is as follows: learning time Δt = 1.5 hours, feedback data: K2 correct answers = 85% (meeting the standard), C2 improved from 5.0 to 6.3 (an improvement of 1.3).

[0164] Reward Calculation:

[0165] Short-term reward = 0.5 × (1.3 / 1.5) = 0.5 × 0.87 ≈ 0.43; Long-term reward (subsequent assessment P = 1) → 0.5 × 1 = 0.5.

[0166] The total reward R2 = 0.43 + 0.5 = 0.93.

[0167] The Q value is iterated again: Q(S1,B)=3.6+0.1×(0.93+0.9×maxQ(S2,A')-3.6)≈3.6+0.05=3.65.

[0168] 6. Results of multiple iterations

[0169] Table 7. Results of Multiple Iterations

[0170] index initial value Value after 3 rounds Improvement rate State space S [5.5,4.2,6.8;0.5,0.3,0.6,0.7] [6.2,7.5,6.8;0.6,0.8,0.6,0.7] - Course B Recommendation Count 1 3 K2's weaknesses continue to worsen Job performance evaluation pass rate 60% 85% 41.7%

[0171] (ii) A safety training method based on a dynamic weight optimization model for capability assessment is provided.

[0172] The method steps of this invention include: (1) constructing a three-dimensional graph model of competence-knowledge-course, wherein the competence dimension is decomposed and weighted based on the job competency model, the knowledge dimension uses knowledge graph technology to construct a knowledge point network, and the course dimension annotates the training resource library with metadata;

[0173] (2) Based on student behavior data, the subjective and objective weights are dynamically adjusted. The initial weights are determined by establishing a capability-knowledge mapping matrix and a knowledge-course mapping relationship. The subjective and objective weight fusion formula is then used. Perform weight optimization, where The decay rate increases exponentially with the amount of training data.

[0174] (3) A personalized course recommendation sequence is generated by reinforcement learning algorithm. The state space is composed of ability assessment vector and knowledge mastery vector, the course resource set is the action space, and the recommendation strategy is constructed using short-term learning efficiency and long-term job assessment pass rate as reward functions.

[0175] Example 4:

[0176] This document outlines a safety training plan for reactor operators at a certain company. The basic requirement is a reactor operator position (involving high-temperature and high-pressure equipment, requiring safe operation and emergency response capabilities). The trainees are 10 newly hired employees (numbered 1-10) from the company. The initial job performance assessment pass rate is 60%, with the core objective of increasing the pass rate to over 85% within three months. The method described below follows specific steps.

[0177] 1. Three-dimensional atlas model

[0178] a. Competency Dimension (Based on Job Competency Model Decomposition)

[0179] Table 8 Examples of Capability Dimensions

[0180] Capability Dimension (Ci) Decomposition Indicators Initial weights (quantized values) C1: Safe Operation Capability Standardized operation proficiency 5.0 / 10 (Medium level) C2: Emergency Response Capability Accident response speed 4.2 / 10 (Slightly low) C3: Equipment maintenance capability Troubleshooting accuracy 6.5 / 10 (Good)

[0181] b. Knowledge Dimension (Knowledge Graph Construction)

[0182] Table 9 Examples of Knowledge Dimensions

[0183] Knowledge Unit (Kj) Relationship Capability Dimension Initial mastery level (0-1) K1: Reactor Pressure Control Specifications C1: Safe Operation Capability 0.5 (normal) K2: Emergency Response Procedures for Hazardous Chemical Spills C2: Emergency Response Capability 0.3 (Weak) K3: Safety Valve Calibration Standard C3: Equipment maintenance capability 0.6 (Medium) K4: Equipment Inspection Cycle Specifications C3: Equipment maintenance capability 0.7 (Good)

[0184] c. Course Dimension (Metadata Annotation)

[0185] Table 10 Examples of Course Dimensions

[0186] Course Resources (Action A) Metadata annotation (covering knowledge units) Duration (hours) A1: Safety Operating Procedures for Reactors Coverage K1 (weight 0.9) 2 A2: Emergency Drill for Hazardous Chemical Spills Coverage K2 (weight 0.95) 3 A3: Safety Valve Calibration and Maintenance Coverage K3 (weight 0.85) 2.5

[0187] 2. Dynamically adjust subjective and objective weights (W) final calculate)

[0188] (1) Determining the initial weights

[0189] Expert weight (W) expert ): Normalized scores from 5 senior engineers (1-5 points).

[0190] Table 11 Example of initial weight determination data

[0191] Ability-Knowledge Mapping <![CDATA[W expert ]]> Knowledge-Course Mapping <![CDATA[W expert ]]> C1-K1 0.8 K1-A1 0.9 C2-K2 0.9 K2-A2 0.95 C3-K3 0.7 K3-A3 0.85

[0192] Data weights (W) data ): Calculated based on data from 300 historical students (answer accuracy rate, learning time).

[0193] Table 12 Examples of Data Weights

[0194] Capability-Knowledge Mapping <![CDATA[W data ]]> Knowledge-Course Mapping <![CDATA[W data ]]> C1-K1 0.75 K1-A1 0.88 C2-K2 0.85 K2-A2 0.92 C3-K3 0.65 K3-A3 0.82

[0195] (2) Weight fusion

[0196] Parameter settings: λ=0.1 (attenuation coefficient), number of iterations t=5 (initial stage, small amount of data); ,1- .

[0197] Here is a calculation example:

[0198] C2-K2 .

[0199] K2-A2 .

[0200] (3) Adjustment after data accumulation (t=30, sufficient data): C2-K2 (More data-dependent).

[0201] 3. Reinforcement learning generates recommendation sequences

[0202] a. State space S and action space A.

[0203] Initial state S0 (Student 1 data):

[0204] S0=[C1=5.0,C2=4.2,C3=6.5;K1=0.5,K2=0.3,K3=0.6].

[0205] Action space A: {A1,A2,A3}.

[0206] b. Calculate the reward function R (ω1=0.6, ω2=0.4).

[0207] Data after the first A2 study:

[0208] Before learning: C 2old =4.2, after learning: C 2new =5.8 (an improvement of 1.6);

[0209] Learning duration Δt = 3 hours, short-term efficiency = (1.6) / 3 ≈ 0.53;

[0210] The initial pass rate for the job assessment was P = 0.6.

[0211] R1=0.6×0.53+0.4×0.6=0.318+0.24=0.558.

[0212] Data after the second A2 study session (recommended again due to low K2 proficiency):

[0213] C 2new =6.9 (increase of 1.1), Δt=2.5 hours, short-term efficiency=1.1 / 2.5=0.44; P=0.75 (after the first assessment)

[0214] R2=0.6×0.44+0.4×0.75=0.264+0.3=0.564.

[0215] 4. Q-score iteration and recommendation strategy optimization

[0216] Q-value update formula: Q(S,A)=Q(S,A)+α[R+γ×maxQ(S',A')-Q(S,A)] (α=0.1, γ=0.9).

[0217] Initially, Q(S0,A2) = 3.2. After the update, Q(S0,A2) = 3.2 + 0.1 × (0.558 + 0.9 × 3.5 - 3.2) = 3.2 + 0.071 = 3.271 (maxQ(S',A') is the optimal Q value after learning).

[0218] Recommended sequence generation:

[0219] Round 1: Q(S0,A2)=3.271 highest → Recommend A2.

[0220] Second round: State S1=[5.0,5.8,6.5;0.5,0.6,0.6], Q(S1,A1)=3.1, Q(S1,A2)=3.3 → Continue to recommend A2.

[0221] Round 3: State S2=[5.2,6.9,6.5;0.55,0.8,0.6], Q(S2,A1)=3.4 (K1 becomes the new weak link) → Recommend A1. Implementation results are shown in the table below.

[0222] Table 13 Implementation Results

[0223] index initial value Final value Improvement rate <![CDATA[Average of Competency Assessment (C1 - C3)]]> 5.23 7.85 50.1% <![CDATA[Mean knowledge mastery level (K1-K3)]]> 0.47 0.82 74.5% Job performance evaluation pass rate (P) 60% 88% 46.7%

[0224] (iii) A safety training system based on an evaluation dynamic weight optimization model and a computer medium are provided.

[0225] See Figures 4-5As shown, this invention presents a computer medium based on the aforementioned security training model, method, and system. In practice, it manifests as a cloud-based or locally deployed computing platform, employing a combination of big data and cloud computing. It features a five-layer architecture, including a user layer, application layer, data layer, service layer, and infrastructure layer. The application layer integrates functional modules such as the platform homepage, basic information management, training services, training activities, points management, and system management, providing users with comprehensive training services.

[0226] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A safety training model based on a dynamic weight optimization model of capability assessment, characterized in that: The model comprises a capability evaluation module, a dynamic weight calculation module, a reinforcement learning recommendation module and a real-time feedback module; The capability evaluation module is configured to extract basic capability, professional capability and management capability three-dimensional indexes required by the employee's post based on a post competency model, generate a personalized capability evaluation vector in combination with employee context characteristics, and the capability evaluation vector contains label levels and initial weights of each capability dimension; The dynamic weight calculation module is configured to build a "capability-knowledge-course" three-dimensional correlation relationship, and perform weight optimization and adjustment, wherein the weight optimization comprises establishing a capability-knowledge mapping matrix, a knowledge-course mapping relationship and a subjective and objective weight fusion; The reinforcement learning recommendation module takes the ability evaluation vector and the knowledge mastery vector as a state space S , the course resource set is an action space, and the short-term learning efficiency ratio of time length to mastery degree and the long-term post assessment pass rate are reward functions R , a recommendation strategy is constructed based on a reinforcement learning framework and a policy gradient algorithm, and a recommended sequence is dynamically optimized through Q value iteration The real-time feedback module is configured to collect student behavior data and context characteristics in real time, and trigger a set mechanism; The model structures the correlation of post competency, professional knowledge system and training course resources through a three-dimensional graph, and forms a dynamically evolving capability training path; The capability evaluation module selects a structural equation model analysis method for index screening, uses a structural equation model to verify the model of the safety capability of the practitioner, indirectly reflects the latent variables that cannot be observed originally, finds the relationship between the latent variables, reflects the relationship between the manifest variables and the latent variables, sets evaluation indexes, and establishes a relationship model between variables; The structural equation model comprises a measurement model and a structure model, and specifically comprises: （1） （2） （3） In equations (1)-(3), In the model, it is represented as the first... j One exogenous manifest variable, Represented as the first j One endogenous explicit variable Represented as the first i An exogenous latent variable, Represented as the first i , One endogenous latent variable, Represented as exogenous manifest variables In exogenous latent variables The factor loading coefficient matrix on, Λy ij Indicates endogenous manifest variables In endogenous latent variables Factor loading coefficient matrix on Represented as exogenous manifest variables The error; Representing endogenous latent variables and The path coefficient matrix between them; Representing exogenous latent variables Endogenous latent variables The impact; Represents the residual term; Indicates endogenous manifest variables The error; The exogenous manifest variables comprise test scores, training attendance rates, simulation operation scores and fault handling speeds; the endogenous manifest variables comprise emergency drill scores, accident handling scores, safety behavior self-evaluation, accident rates and safety inspection scores; the endogenous latent variables comprise emergency capability, safety attitude and safety performance; and the exogenous latent variables comprise safety knowledge and operation skills.

2. The safety training model based on the dynamic weight optimization model of capability evaluation according to claim 1, wherein: The specific way of building the "capability-knowledge-course" three-dimensional correlation relationship comprises: (1) establishing a capability-knowledge mapping matrix, determining the initial correlation weight of the capability dimension and the knowledge unit based on expert knowledge graph and historical training data mining; (2) establishing a knowledge-course mapping relationship, determining the coverage weight of the knowledge unit and the course resource through course outline analysis, test knowledge point marking and learning effect backstepping; (3) Fusion of subjective and objective weights, set dynamic adjustment formula , wherein the subjective weight Based on the analytic hierarchy process AHP Or Delphi method to determine, the objective weight Based on the student's answer accuracy, learning time, simulation operation score of behavior through data correlation algorithm dynamic calculation, weight adjustment factor Exponential decay with the increase of training data; decay factor , t The number of iterations, The attenuation coefficient.

3. The safety training model based on the dynamic weight optimization model of capability evaluation according to claim 1, characterized in that: In the reinforcement learning recommendation module, wherein the state space , represents the evaluation value of the first i, m ability dimension, represents the mastery degree of the first j, n knowledge unit; the reward function , , is a weight coefficient, 、 respectively, the ability dimension evaluation value before learning and after learning, is the learning duration, and the post examination pass rate.

4. The safety training model based on the dynamic weight optimization model of capability evaluation according to claim 1, wherein: The context features include at least one of tenure, job type, post type, post risk level, and history training record; and the reinforcement learning framework includes deep Q network DQN, policy gradient algorithm Policy Gradient at least one; and the behavior data includes at least one of correct answer rate, learning duration, interactive operation score, and video review times.

5. The safety training model based on the dynamic weight optimization model of capability evaluation according to claim 1, characterized in that: The set mechanism comprises: (1) Knowledge node importance update, real-time collection of student operation behavior data, assessment performance data, and knowledge mastery data, using the collected data to dynamically adjust the node weights in the knowledge point network through PageRank algorithm, update the importance of knowledge nodes, establish a safety capability evaluation system for chemical industry practitioners, and based on the dynamic weight optimization model of the safety training model, comprehensively consider the factors of employee psychological quality, business technical ability, and safety operation skill, provide basis for safety evaluation, pre-training investigation, and post-training effect verification; (2) Abnormal knowledge node detection, when the standard deviation of the student's correct answer rate is greater than 0.3, or the set detection knowledge point is continuously answered incorrectly N threshold, triggering weight adjustment and course resource association reconstruction; (3) weight adjustment factor self-adaption, dynamically adjusting the subjective and objective weight fusion proportion according to the data sample size, preferentially using the expert experience weight when the initial sample size, and gradually increasing the proportion of the data-driven weight with the accumulation of data.

6. A safety training method based on a dynamic weight optimization model of capability evaluation, characterized in that: The safety training method is training based on the safety training model of the capability evaluation dynamic weight optimization model according to any one of claims 1-5, and the safety training method comprises (1) building a capability-knowledge-course three-dimensional graph model, wherein the capability dimension is decomposed and quantified based on the post competency model, the knowledge dimension is constructed by using the knowledge graph technology to construct a knowledge point network, and the course dimension is metadata annotated to the training resource library; (2) Dynamically adjust the subjective and objective weights based on the student behavior data, determine the initial weights by establishing the ability-knowledge mapping matrix and the knowledge-course mapping relationship, and use the subjective and objective weight fusion formula Optimize the weights, wherein Exponential decay with the increase of training data (3) Generating a personalized course recommendation sequence through a reinforcement learning algorithm, taking the ability evaluation vector and the knowledge mastery vector as the state space, the course resource set as the action space, and the short-term learning efficiency and the long-term post examination pass rate as the reward function to construct the recommendation strategy.

7. A safety training system based on capability assessment dynamic weight optimization model, characterized in that: The system works based on the safety training model of any one of claims 1-5, including an initialization phase, a running phase, and an optimization iteration phase, specifically: (1) Initialization phase: (1.1) The capability assessment module inputs the individualized capability assessment vector containing the grades of each capability dimension and the initial weights, and outputs the individualized capability assessment vector of the grades of the dimension labels and the initial weights generated by the post competency model and the optimized "capability-knowledge-course" mapping weights ; (1.2) Screen out exogenous and endogenous variables through structural equation modeling; (1.3) The capability evaluation module delivers each capability dimension evaluation value in the capability evaluation vector to the reinforcement learning recommendation module with and each knowledge unit initial mastery degree （ , ,…,K n ） , as the initial value of the reinforcement learning state space ​ (1.4) The dynamic weight optimization module sends the constructed initial weights of the "ability-knowledge-course" three-dimensional correlation to the reinforcement learning recommendation module, including the initial weight data of the ability-knowledge mapping matrix and the knowledge-course mapping relationship; (2) Running phase: (2.1) The reinforcement learning recommendation module calls the latest "ability-knowledge-course" three-dimensional correlation weight data through the dynamic weight optimization module in real time to calculate the recommendation strategy; (2.2) The dynamic weight optimization module returns the fusion of the subjective and objective weights according to the recommendation of the reinforcement learning recommendation module computed fusion of the subjective and objective weights wherein , is a decay coefficient; (2.3) The reinforcement learning recommendation module pushes the personalized course recommendation sequence generated based on the reinforcement learning framework and the policy gradient algorithm to the real-time feedback module, and the sequence is Q value iteration dynamic optimization; (2.4) The real-time feedback module transmits the student behavior data and characteristics to the reinforcement learning recommendation module in real time; The real-time feedback module transmits the student behavior data and examination data to the ability evaluation module to trigger the update of the ability evaluation vector and the knowledge mastery vector; The real-time feedback module sends student action data, assessment score data, and knowledge mastery data to the dynamic weight optimization module, triggering updates to the importance of knowledge nodes. PageRank The algorithm adjusts the weights of knowledge point network nodes; when the abnormal knowledge node detection conditions are met, the standard deviation of the answer accuracy is >0.3 or the set detection knowledge points are continuous. N When the error rate in answering questions exceeds the threshold, a weight adjustment and a reconstruction of the course resource association relationship are triggered. (3) Optimization iteration phase: (3.1) The updated ability evaluation vector containing the latest evaluation value of each ability dimension and the knowledge mastery vector are sent to the dynamic weight optimization module to trigger weight optimization adjustment; (3.2) The dynamic weight optimization module pushes the optimized "ability-knowledge-course" three-dimensional correlation weight data to the reinforcement learning recommendation module to update the recommendation strategy; (3.3) The capability evaluation module delivers the updated capability evaluation vector and the knowledge mastery degree vector to the reinforcement learning recommendation module, and updates the reinforcement learning state space S ; (3.4) Real-time feedback module: provide the ability dimension evaluation value of the student before and after learning 、 , learning duration ∆t and post assessment pass rate P , used to calculate the reward function , the reinforcement learning recommendation module optimizes the recommended strategy based on this Q value iteration (3.5) Dynamic weight optimization module: synchronize the change information of weight adjustment factor a, and the real-time feedback module adjusts the data acquisition frequency and the threshold parameters of the trigger mechanism accordingly; when it is detected that the student's ability improves significantly after learning a certain type of course, the change amplitude of the ability evaluation value exceeds the set threshold value γ or has no obvious improvement or is lower than the threshold value δ , the real-time feedback module sends course effectiveness evaluation data to assist the dynamic weight optimization module in deep calibration of the "ability-knowledge-course" correlation.

8. A computer medium, characterized by The computer medium stores programs or instructions, and the computer programs or instructions are executed by the processor to realize the steps of the safety training method based on the ability evaluation dynamic weight optimization model of claim 6.

Images