An adaptive learning method for controller simulation training

Abstract

The application discloses an adaptive learning method for controller simulation training, and belongs to the technical field of research on an air traffic control teaching training system, and specifically relates to solving the problems in the background art that a unified teaching material and a unified teaching mode are used in traditional controller simulation training teaching, it is difficult to stimulate the learning initiative and innovative ability of students, and subjective scoring of teachers is mainly relied on for controller training evaluation, and objectivity is lacked, and the adaptive learning method for controller simulation training is established, main features of controllers are acquired by collecting basic information of the controllers, and a portrait is constructed; training data of the controllers are collected in real time and stored; a learning state of a current knowledge point of the controllers is evaluated through a Bayesian network and deviation degree analysis, unmastered content in a training process is judged according to a diagnosis evaluation result, training suggestions, learning routes and learning content are targetedly recommended for reference and improvement of the controllers, and training efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of research and technology of air traffic control teaching and training systems, and specifically relates to an adaptive learning method for controller simulation training. Background Technology

[0002] With the increasing development of technologies such as cloud computing, big data, and artificial intelligence, it has become a trend to use various artificial intelligence algorithms and big data technologies to analyze the factors affecting the training effect of air traffic controllers and to continuously improve and optimize training methods.

[0003] As air traffic controllers, they need to master a variety of skills and be highly proficient in them. Currently, in front-line air traffic control units in my country, the assessment of controllers' work ability and command level is based on the relevant training management rules and unified teaching materials issued by the Civil Aviation Administration of China, and on the assessment element checklist. The assessment instructors manually score and evaluate the controllers' instructions, control awareness, decision-making level and other aspects during the operation.

[0004] However, the traditional controller simulation training uses uniform teaching materials and a uniform teaching model, which makes it difficult to stimulate trainees' learning initiative and innovation ability, ultimately resulting in mediocre training effects for airport controllers. In addition, the traditional training method mainly relies on the instructor's subjective scoring for controller training evaluation, which lacks objectivity.

[0005] Traditional methods for improving learning and training efficiency mainly rely on instructors' experience and students' own exploration. They lack objective analysis and summarization based on a large amount of training data, fail to objectively assess students' learning characteristics and their mastery of knowledge, and do not conduct targeted gap-filling exercises. This results in repetitive practice of some content or a lack of practice of other content, ultimately leading to low learning and training efficiency.

[0006] Therefore, there is an urgent need to design an adaptive learning method for controller simulation training to solve the above problems. Summary of the Invention

[0007] To address the aforementioned issues, this application establishes an adaptive learning method for controller simulation training. This method collects basic information about controllers to obtain their main characteristics and constructs a profile; it collects and stores controller training data in real time; it evaluates the controller's current learning status of knowledge points through Bayesian networks and deviation analysis; based on the diagnostic evaluation results, it identifies content that has not been mastered during training and provides targeted training suggestions, learning paths, and learning content for controllers to refer to and improve their training, thereby enhancing training efficiency.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] An adaptive learning method for controller simulation training, characterized in that: the method includes:

[0010] Step 1. Create a profile of the controller;

[0011] Step 2. Collect and store the controller's basic information and training information from Step 1;

[0012] Step 3. Extract the controller's basic information and training information from Step 2, and use them as nodes to construct a Bayesian network;

[0013] Step 4. Obtain the relationships between the Bayesian network layers from Step 3;

[0014] Step 5. Bayesian network parameter update: When new data is generated at a node, repeat steps 3 and 4 to update the probabilities of each node in the Bayesian network in real time, thus forming an adaptive network parameter update.

[0015] Step 6. Based on the observed variables, historical training scores, and Bayesian network node content, use the variance formula to perform deviation analysis and statistically analyze the controller's mastery of the knowledge points at each node.

[0016] Step 7: Combining the unmastered node content identified in Step 6 with the controller profile from Step 1, adaptively recommend the next node's knowledge points and learning path, as well as recommend targeted training subjects.

[0017] Preferably, the specific process of constructing the controller profile in step one includes:

[0018] 101. Collect basic data from air traffic controllers;

[0019] 102. Determine the controller's learning style based on their learning behavior and habits, and assess their current level of understanding of relevant knowledge by analyzing the duration of subject training, the number of training sessions, and the number of times instructions were modified.

[0020] 103. Extract the controller's feature data from processes 101 and 102, and select key features through expert judgment to serve as controller profile labels and construct a profile;

[0021] 104. As controllers' cognitive and training levels improve, their profiles will be automatically updated periodically.

[0022] Preferably, the controller's basic information in step 2 includes: controller's gender, age, education, major, and years of experience in controller management; training information includes: controller's training subjects, number of training sessions, training time, training score, and voice data, safety intervals, process alarms, and deviation reminders during training.

[0023] Preferably, the specific process of constructing the Bayesian network in step 3 includes:

[0024] 301. Construct hierarchical relationships;

[0025] Based on the controllers' historical training information collected in step 2, all controllers' simulation training knowledge points are divided into three levels: A, B, and C.

[0026] Layer C represents the foundation layer, which contains only basic knowledge;

[0027] Layer B represents module knowledge that is higher than Level C and contains multiple basic knowledge modules;

[0028] Layer A represents the overall training effect of the simulation training, and is also denoted as the observed variable;

[0029] 302. Constructing a Bayesian network structure:

[0030] The knowledge of nodes at various levels forms a Bayesian network. The nodes in the Bayesian network include: A-level nodes, B-level nodes, and C-level nodes. m represents the state of each level of nodes that the controller knows, and they are represented by Y / N respectively.

[0031] Each node in level C represents the knowledge point at the lowest level;

[0032] B-level nodes represent knowledge points at a higher level than C-level nodes;

[0033] Level A nodes represent the final simulated training results and are also denoted as observed variables;

[0034] Y indicates mastery;

[0035] N indicates that the knowledge is not yet available.

[0036] Preferably, the specific process for obtaining the association relationships between the Bayesian network layers in step 4 includes:

[0037] 401. Sample Collection:

[0038] The controller examination subjects are divided into three stages: primary, intermediate and advanced. N sets of training data are randomly selected from each stage and the operation is repeated m times to form the final sample S.

[0039] 402. Calculate the prior probabilities of each node in a Bayesian network using the following formula:

[0040]

[0041]

[0042] P(m) represents the prior knowledge probability of each node;

[0043] n0 represents the number of all sample groups in sample S where m = Y;

[0044] n1 represents the number of all sample groups in sample S where m = N;

[0045] 403. Calculate the conditional probabilities of nodes at each level in a Bayesian network. The calculation formula is as follows:

[0046]

[0047]

[0048]

[0049]

[0050] S1 represents the number of sample groups in sample S where the state of node B is "Y";

[0051] n1 is the number of sample groups in S1 where the state of node B is "Y" and the state of node A is "Y".

[0052] n2 is the number of sample groups in S1 where node B is in the state of "Y" and node A is in the state of "N".

[0053] S2 represents the number of sample groups in sample S where the state of node B is "N";

[0054] n3 is the number of sample groups in S1 where node B is in state "N" and node A is in state "Y".

[0055] n4 is the number of sample groups in S1 where the state of node B is "N" and the state of node A is "N".

[0056] 404. Calculate the joint probability among the nodes in a Bayesian network;

[0057] Calculate the joint probability P(x_1,…,x_n) among the nodes in the Bayesian network according to the law of total probability.

[0058] P(x_1,…,x_n)=P(x_(n-1)|x_1,…,x_(n-2))P(x_n|x_1,…,x_(n-1))P(x_1,…,x_(n-2))…P(x_2|x_1)P(x_1)

[0059] x represents the probability value of each node. The probability of satisfying the requirements of all nodes can be viewed through the joint probability.

[0060] 405. Determining the correlation between nodes and observed variables in a Bayesian network:

[0061] Modify the prior probability or conditional probability of the node, and check the degree of influence on the final observed object A. The greater the degree of influence, the closer the current node is related to the observed object.

[0062] Preferably, in step 6, the variance formula is used for deviation analysis. The specific process of statistically analyzing the controller's knowledge of each node is as follows:

[0063] 601. Use the variance formula to judge the stability φ of the controller's mastery of the scores of each node in the Bayesian network;

[0064]

[0065] M represents the average score of each node in the Bayesian network;

[0066] n represents the total number of training times;

[0067] x represents the score of the node content each time, and the value result is 0 or 100;

[0068] The value range of the node score stability φ is as follows:

[0069]

[0070] When b = 0, it means the error is 0, and the result is recorded as T;

[0071] When 0 < b ≤ 10, it means the score stability is within the controllable range, and the result is recorded as T;

[0072] When b > 10, it means the score is unstable and exceeds the controllable range, and the result is recorded as F;

[0073] 602. Judge the mastery situation S of the knowledge points of each node in the Bayesian network. The judgment method includes:

[0074] Calculate the controller's mastery of the knowledge points of each node through the following formula:

[0075]

[0076] N represents the state of "not mastered" for the mastery of node knowledge points;

[0077] Y represents the state of "mastered" for the mastery of node knowledge points;

[0078] φ represents the stability of the current node score;

[0079] M represents the average score of the current node;

[0080] When the average score M of the node is greater than or equal to 60 and φ is T, the trainee's mastery state of the knowledge points of the current node is "mastered";

[0081] When the average score M of a node is greater than or equal to 60 and φ is F, the student's mastery status of the knowledge point of the current node is "not mastered";

[0082] When the average score M of a node is less than 60, the student's mastery status of the knowledge point in the current node is "not mastered".

[0083] The beneficial effects of this invention are:

[0084] This invention uses Bayesian networks and deviation analysis to evaluate the current learning status of controllers' knowledge points. Based on the evaluation results, it identifies the content that controllers have not mastered during training and provides targeted training suggestions, learning routes, and learning content for controllers to refer to and improve their training, thereby increasing training efficiency. Attached Figure Description

[0085] Figure 1 This is a flowchart of the present invention;

[0086] Figure 2 This is a flowchart of the controller portrait component of the present invention;

[0087] Figure 3 This is a flowchart of the controller training information collection process of the present invention;

[0088] Figure 4 This is a flowchart of the intelligent diagnosis and adaptive recommendation process for controller simulation training in this invention;

[0089] Figure 5 This is a Bayesian network structure diagram of the knowledge variables from controller simulation training in this invention;

[0090] Figure 6 Reference Case of Bayesian Network Based on Knowledge Variables from Controller Simulation Training; Detailed Implementation

[0091] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0092] Example 1:

[0093] See attached document Figure 1-6 As shown, an adaptive learning method for controller simulation training includes the following steps:

[0094] Step 1. Create a profile of the controller;

[0095] 101. Collect basic data on controllers, including years of experience, gender, age, education, major, etc.

[0096] 102. Determine the controller's learning style based on learning behavior and habits, and assess the controller's current level of knowledge by analyzing information such as subject training duration, number of training sessions, and number of instruction modifications.

[0097] 103. Extract the controller's feature data through steps 101 and 102, and have key features judged and selected by experts to serve as controller profile tags to construct profiles, such as the user's learning characteristics and the strength of their learning ability.

[0098] 104. As controllers' cognitive and training levels improve, their profiles will be automatically updated periodically.

[0099] Step 2. Collect and store basic information and training information for controllers;

[0100] 201: Collect basic training data of controllers in real time, including information such as years of experience in controllership, gender, age, education, and major.

[0101] 202: Real-time collection of controller training process data, including voice data, safety intervals, process alarms, deviation alerts, etc. Step 3. Extract the controller's basic information and training information from Step 2, and use them as nodes to construct a Bayesian network;

[0102] Step 3. Extract the controller's basic information and training information from Step 2, and use them as nodes to construct a Bayesian network;

[0103] 301. Construct hierarchical relationships;

[0104] Based on the controllers' historical training information collected in step 2, all controllers' simulation training knowledge points are divided into three levels: A, B, and C.

[0105] Layer C represents the foundation layer, which contains only basic knowledge;

[0106] Layer B represents module knowledge that is higher than Level C and contains multiple basic knowledge modules;

[0107] Layer A represents the overall training effect of the simulation training, and is also denoted as the observed variable;

[0108] 302. Constructing a Bayesian network structure:

[0109] The knowledge of nodes at various levels forms a Bayesian network. The nodes in the Bayesian network include: A-level nodes, B-level nodes, and C-level nodes. m represents the state of each level of nodes that the controller knows, and they are represented by Y / N respectively.

[0110] Each node in level C represents the knowledge point at the lowest level;

[0111] B-level nodes represent knowledge points at a higher level than C-level nodes;

[0112] Level A nodes represent the final simulated training results and are also denoted as observed variables;

[0113] Y indicates mastery;

[0114] N indicates that the knowledge is not yet available.

[0115] Step 4. Obtain the relationships between the Bayesian network layers in Step 3, that is, train the Bayesian network parameters, which further means calculating the prior probability and conditional probability of each level of node;

[0116] 401. Sample Collection:

[0117] The controller examination subjects are divided into three stages: primary, intermediate and advanced. N sets of training data are randomly selected from each stage and the operation is repeated m times to form the final sample S.

[0118] 402. Calculate the prior probabilities of each node in a Bayesian network using the following formula:

[0119]

[0120]

[0121] P(m) represents the prior knowledge probability of each node;

[0122] n0 represents the number of all sample groups in sample S where m = Y;

[0123] n1 represents the number of all sample groups in sample S where m = N;

[0124] 403. Calculate the conditional probabilities of nodes at each level in a Bayesian network. The formula is as follows:

[0125]

[0126]

[0127]

[0128]

[0129] S1 represents the number of sample groups in sample S where the state of node B is "Y";

[0130] n1 is the number of sample groups in S1 where the state of node B is "Y" and the state of node A is "Y".

[0131] n2 is the number of sample groups in S1 where node B is in the state of "Y" and node A is in the state of "N".

[0132] S2 represents the number of sample groups in sample S where the state of node B is "N";

[0133] n3 is the number of sample groups in S1 where node B is in state "N" and node A is in state "Y".

[0134] n4 is the number of sample groups in S1 where the state of node B is "N" and the state of node A is "N".

[0135] For example, in 100 training data points, 20 C_1 node contents have been mastered, and 10 of the 20 C_1 node contents have been mastered, then P(B_1=Y|C_1=Y)=0.1;

[0136] 404. Calculate the joint probability among the nodes in a Bayesian network;

[0137] Calculate the joint probability P(x_1, ..., x_n) among the nodes in the Bayesian network according to the law of total probability.

[0138] P(x_1,…,x_n)=P(x_(n-1)|x_1,…,x_(n-2))P(x_n|x_1,…,x_(n-1))P(x_1,…,x_(n-2))…P(x_2|x_1)P(x_1)

[0139] x represents the probability value of each node. The probability of satisfying the requirements of all nodes can be viewed through the joint probability.

[0140] 405. Determining the correlation between nodes and observed variables in a Bayesian network:

[0141] Modify the prior or conditional probability of a node and observe its impact on the final observed object A. The greater the impact, the stronger the correlation between the current node and the observed object. Refer to a real-world example for details. Figure 6 Here, the observed object A is an instance of the observed variable;

[0142] Step 5. Bayesian network parameter update: When new data is generated at a node, repeat steps 3 and 4 to update the probabilities of each node in the Bayesian network and other nodes at all levels in real time, thus forming an adaptive network parameter update.

[0143] Step 6. Based on the observed variables, historical training scores, and Bayesian network node content, use the variance formula to perform deviation analysis and statistically analyze the controller's mastery of the knowledge points at each node.

[0144] 601. Use the variance formula to judge the stability φ of the mastery of the scores of each node in the Bayesian network;

[0145]

[0146] $M$ represents the average score of each node in the Bayesian network;

[0147] $n$ represents the total number of training times;

[0148] $x$ represents the score of the node content each time, and the value result is 0 or 100;

[0149] The value range of the node performance stability $\varphi$ is as follows:

[0150]

[0151] When $b = 0$, it means the error is 0, and the result is recorded as $T$;

[0152] When $0 < b \leq 10$, it means the performance stability is within the controllable range, and the result is recorded as $T$;

[0153] When $b > 10$, it means the performance is unstable and exceeds the controllable range, and the result is recorded as $F$;

[0154] 602. Judge the mastery situation $S$ of each node's knowledge points in the Bayesian network. The judgment method includes:

[0155] Calculate the mastery situation of each node's knowledge points of the controller through the following formula:

[0156]

[0157] $N$ represents that the mastery status of the node's knowledge point is "not mastered";

[0158] $Y$ represents that the mastery status of the node's knowledge point is "mastered";

[0159] $\varphi$ represents the current node performance stability;

[0160] $M$ represents the average score of the current node;

[0161] When the average score $M$ of the node is greater than or equal to 60 and $\varphi$ is $T$, the trainee's mastery status of the current node's knowledge point is "mastered";

[0162] When the average score $M$ of the node is greater than or equal to 60 and $\varphi$ is $F$, the trainee's mastery status of the current node's knowledge point is "not mastered";

[0163] When the average score $M$ of the node is less than 60, the trainee's mastery status of the current node's knowledge point is "not mastered".

[0164] Step 7: Combine the unmastered node content statistically in Step 6 and the controller portrait in Step 1 to perform adaptive recommendation of the next node's knowledge point and learning path, as well as targeted training subject recommendation.

[0165] Based on the statistical analysis of unmastered content, targeted training subjects are recommended, along with adaptive recommendations for the next knowledge point and learning path.

[0166] Recommended learning materials for concepts you haven't mastered:

[0167] If a student's current C_3 node knowledge point is "not mastered", C_3 related learning content will be recommended.

[0168] If a student's mastery status for the current B_3 node knowledge point is "not mastered", C_5, which has the strongest correlation in the previous path, will be recommended first, followed by other node knowledge points;

[0169] Recommend targeted training subjects based on user profiles:

[0170] For example, students with poor hands-on skills should be given priority for hands-on practice in subjects that require more training.

[0171] For students who lack theoretical knowledge, we will prioritize recommending theoretical learning materials for the knowledge points they have not yet mastered;

[0172] Students with weak self-learning abilities will be given priority in being recommended to teachers in their respective majors for guidance;

[0173] For students with a weaker foundation, we recommend studying basic theoretical materials or practicing basic computer simulations.

[0174] Step 7: Based on the unmastered node content identified in Step 6, make targeted training subject recommendations, as well as adaptive recommendations for the next node's knowledge points and learning path.

[0175] Example 2:

[0176] Example 2 applies the method in Example 1 to a real control tower flight command simulation system. Based on the system's training knowledge points and simulation training results, this adaptive learning method is used to construct a profile through controller profile construction technology, collect a large amount of historical training process data, and collect and store controller training information using controller training information collection technology.

[0177] Using a Bayesian network, we extract knowledge point variables related to the "comprehensive training effect of simulation training" from the acquired basic information and historical simulation training data. These variables are then used as nodes to construct the Bayesian network. Based on steps 301 to 304, we calculate the degree of association between nodes and statistically analyze the students' mastery of each knowledge point. To simplify the demonstration, only a portion of the knowledge point variables are selected here. Refer to the network structure and probability diagrams of each node as shown below. Figure 6 As shown.

[0178] Modify the conditional probabilities of each node and observe the degree of impact on the "comprehensive training effect of simulation training" of the observed object:

[0179] Through multiple trials, when P(simulated training overall training effect = Y | interval standard = N) =

[0180] When P(Simulated Training Comprehensive Training Effect = Y | Interval Standard = N) = 52.3, the observed object "Simulated Training Comprehensive Training Effect" shows the greatest change. Therefore, the "Interval Standard" module has the greatest correlation with "Simulated Training Comprehensive Training Effect" and can be used as the preferred learning path for "Simulated Training Comprehensive Training Effect".

[0181] Based on the knowledge mastery status of a particular student as statistically analyzed in step 304:

[0182] If your current knowledge of "air-to-ground communication" is "not mastered", then we recommend learning materials related to communication phrases.

[0183] The current knowledge status of the "Emergency Response" node is "not mastered". Priority will be given to recommending the knowledge related to "engine shutdown" which has the greatest correlation in the previous path.

[0184] If the current student's score on "Comprehensive Training Effectiveness in Simulation Training" is low, then learning content related to the interval standard will be recommended first.

[0185] This method can effectively provide precise and personalized services, improving trainees' learning enthusiasm and innovation capabilities. The Bayesian network diagnostic technology and variance deviation analysis employed can effectively and objectively assess controllers' knowledge mastery level, reducing the subjectivity of instructor assessments. This method automatically analyzes data during controller training and automatically adjusts the learning experience for controllers, reducing repetitive training and thus improving training efficiency and level.

[0186] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive learning method for controller simulation training, characterized in that: The method includes: Step 1. Create a profile of the controller; Step 2. Collect and store the controller's basic information and training information from Step 1; Step 3. Extract the controller's basic information and training information from Step 2, and use them as nodes to construct a Bayesian network; Step 4. Obtain the relationships between the Bayesian network layers from Step 3; Step 5. Bayesian network parameter update: When new data is generated at a node, repeat steps 3 and 4 to update the probabilities of each node in the Bayesian network in real time, thus forming an adaptive network parameter update. Step 6. Based on the observed variables, historical training scores, and Bayesian network node content, use the variance formula to perform deviation analysis and statistically analyze the controller's mastery of the knowledge points at each node. Step 7: Combining the unmastered node content identified in Step 6 with the controller profile in Step 1, make adaptive recommendations for the next node's knowledge points and learning paths, as well as targeted training subject recommendations. The specific process of constructing the Bayesian network in step 3 includes:

301. Construct hierarchical relationships; Based on the controllers' historical training information collected in step 2, all controllers' simulation training knowledge points are divided into three levels: A, B, and C. Layer C represents the foundation layer, which contains only basic knowledge; Layer B represents module knowledge that is higher than Level C and contains multiple basic knowledge modules; Layer A represents the overall training effect of the simulation training, and is also denoted as the observed variable; 302. Constructing a Bayesian network structure: The knowledge of nodes at various levels forms a Bayesian network. The nodes in the Bayesian network include: A-level nodes, B-level nodes, and C-level nodes. m represents the state of each level of nodes that the controller knows, and they are represented by Y / N respectively. Each node in level C represents the knowledge point at the lowest level; B-level nodes represent knowledge points at a higher level than C-level nodes; Level A nodes represent the final simulated training results and are also denoted as observed variables; Y indicates mastery; N indicates that it is not known; The specific process of obtaining the association relationships between the Bayesian network layers in step 4 includes:

401. Sample collection: The controller examination subjects are divided into three stages: primary, intermediate and advanced. N training data are randomly selected from each stage and the operation is repeated m times to form the final sample S.

402. Calculate the prior probabilities of each node in a Bayesian network; 403. Calculate the conditional probabilities of nodes at each level in a Bayesian network; 404. Calculate the joint probability among the nodes in a Bayesian network; 405. Determining the correlation between nodes and observed variables in a Bayesian network: Modify the prior probability or conditional probability of the node and check the degree of influence on the final observed object A. The greater the degree of influence, the closer the relationship between the current node and the observed object.

2. The adaptive learning method for controller simulation training according to claim 1, characterized in that: In step 402, the prior probabilities of each node in the Bayesian network are calculated using the following formula: ； ； P(m) represents the prior knowledge probability of each node; n0 represents the number of all sample groups in sample S where m=Y; n1 represents the number of all sample groups in sample S where m=N; In step 403, calculate the conditional probabilities of each level of nodes in the Bayesian network. The calculation formula is as follows: ； ； ； ； S1 represents the number of sample groups in sample S where the state of node B is "Y". n1 is the number of sample groups in S1 where the state of node B is "Y" and the state of node A is "Y". n2 is the number of sample groups in S1 where the state of node B is "Y" and the state of node A is "N". S2 represents the number of sample groups in sample S where the state of node B is "N". n3 is the number of sample groups in S1 where the state of node B is "N" and the state of node A is "Y". n4 is the number of sample groups in S1 where the state of node B is "N" and the state of node A is "N". In step 404, calculate the joint probability P(x_1,…,x_n) between each node in the Bayesian network. The calculation formula is as follows: P(x_1,…,x_n) = P(x_(n-1)|x_1,…,x_(n-2))P(x_n|x_1,…,x_(n-1))P(x_1,…,x_(n-2))…P(x_2|x_1)P(x_1) x represents the probability values of each node. Through the joint probability, the probability that meets the requirements of all nodes can be viewed.

3. The adaptive learning method for controller simulation training according to claim 1, characterized in that: The specific process of constructing the controller portrait in step 1 includes:

101. Collect basic data of the controller.

102. Judge the learning style of the controller according to the learning behavior and learning habits of the controller. Judge the cognitive level of the current controller for relevant knowledge through the information of subject training duration, training times, and modified instruction times.

103. Extract the characteristic data of the controller from process 101 and process 102, and after expert judgment and screening, select the key features as the controller portrait labels to construct the portrait.

104. As the cognitive level and training level of the controller improve, automatically update the controller portrait regularly.

4. The adaptive learning method for controller simulation training according to claim 1, characterized in that: The basic information of the controller in step 2 includes: controller gender, age, education level, major, and controller years of service information; the training information includes: controller training subjects, training times, training time, training scores, as well as voice data, safety intervals, process warnings, and deviation reminder information during the training process.

5. The adaptive learning method for controller simulation training according to claim 1, characterized in that: In step 6, use the variance formula for deviation analysis. The specific process of counting the mastery of knowledge points of each node by the controller is as follows:

601. Using the variance formula, assess the stability of the performance of each node in a Bayesian network. Make a judgment; ； M represents the average score of each node in the Bayesian network. n represents the total number of trainings. x represents the score of each node content each time, and the value result is 0 or 100. Node performance stability The range of values ​​is as follows: ； When b = 0, it means the error is 0, and the result is recorded as T. When 0 < b ≤ 10, it means the performance stability is within the controllable range, and the result is recorded as T. When b > 10, it means the performance is unstable and exceeds the controllable range, and the result is recorded as F.

602. Judge the mastery situation S of the knowledge points of each node in the Bayesian network. The judgment method includes: Calculate the mastery of the knowledge points of each node of the controller through the following formula: ； N represents the state of "not mastered" for the mastery of node knowledge points. Y represents the state of "mastered" for the mastery of node knowledge points. This indicates the stability of the current node's performance; M represents the average score of the current node. When the node average score M is greater than or equal to 60, and When T is the value, the student's mastery status of the knowledge point at the current node is "mastered"; When the node average score M is greater than or equal to 60, and When the value is F, the student's mastery status for the current knowledge point is "not mastered"; When the average score M of the node is less than 60, the trainee's mastery state of the knowledge points of the current node is "not mastered".

Images