AcT-R-based personnel cognitive behavior modeling method and cognitive load calculation method

Abstract

The invention relates to the technical field of cognitive system application, and particularly discloses an ACT-R-based personnel cognitive behavior modeling method and a cognitive load calculation method.The cognitive load calculation method comprises the steps that on the basis of the relation between module activation time and weight in an ACT-R cognitive system, the instantaneous cognitive load in the task execution process is calculated; during model operation, according to the task process and the activation time of the knowledge base calling module, the instantaneous cognitive load of each module is dynamically calculated; integrating the instantaneous cognitive load in the task execution process and the instantaneous cognitive load of each module to obtain the personnel cognitive load; a weight distribution and error correction mechanism is used to realize personnel cognitive load quantitative evaluation of different task complexity and personnel states; the cognitive load calculation formula embedded in the cognitive model can calculate the cognitive load generated when each module is called, and reference is further provided for personnel cognitive behavior analysis and work performance improvement.

Description

Technical Field

[0001] The present invention belongs to the technical field of cognitive system application, and in particular relates to a personnel cognitive behavior modeling method and a cognitive load calculation method based on ACT-R. Background Art

[0002] The ACT-R (Adaptive Control of Thought–Rational) cognitive system was proposed by Anderson and his team in the 1970s and is still being improved. There are many successful applications based on this cognitive system, such as instructional design, automobile driving model construction, ship steering model construction, air traffic control simulation, etc. ACT-R has a relatively mature declarative knowledge system based on block activation and a procedural knowledge system based on production rules.

[0003] In the ACT-R cognitive system, the human brain is divided into eight cognitive modules according to its functions, namely: vision, hearing, action, language, imagination, goal, declarative knowledge and procedural knowledge modules. Among them, the four modules of vision, hearing, action and language are responsible for interacting with the external environment; the imagination module is responsible for processing imaginary information, such as imagining the consequences of behavior, recalling knowledge, comparing sizes, etc.; the goal module is responsible for setting behavioral goals; the declarative knowledge module is responsible for processing conceptual information; the procedural knowledge module, as the core part of the ACT-R system, is responsible for logical thinking and responding accordingly to the external environment to achieve the goals set by the goal module

[0004] The construction of the knowledge base is the focus of modelers. Knowledge is mainly divided into declarative knowledge and procedural knowledge. Declarative knowledge can encode simple facts, such as Beijing is the capital of China; procedural knowledge is like "IF-THEN", which consists of conditions and operation rules. When the corresponding conditions are met, the corresponding operation is generated. When the conditions are matched and the production is activated, the corresponding action can add or change the declarative knowledge, such as setting a new goal or issuing an action command. The activation level of these blocks is also like human memory, which will decay over time or be enhanced as the activation increases.

[0005] Thanks to the support of psychological theories and the development of related experiments, ACT-R has conducted in-depth research on human's advanced cognitive processes such as learning, memory, decision-making, etc., and is different from traditional psychological qualitative analysis. ACT-R can quantitatively simulate human cognitive behavior and calculate cognitive load, which provides a more convenient way to understand people's cognitive behavior.

[0006] In this regard, the inventors proposed a personnel cognitive behavior modeling method and a cognitive load calculation method based on ACT-R to solve the above problems. Summary of the invention

[0007] The purpose of the present invention is to provide a personnel cognitive behavior modeling method and a cognitive load calculation method based on ACT-R to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The personnel cognitive behavior modeling method based on ACT-R is characterized by comprising the following steps:

[0010] The cognitive process of personnel performing tasks is divided into three parts: perception modeling, decision modeling and control modeling;

[0011] The perception modeling obtains information from the external environment through the perception module, stores the information in the buffer, and processes it by the knowledge module;

[0012] The decision modeling activates production rules based on the information in the buffer, matches the corresponding knowledge base content, and selects applicable rules through a conflict resolution mechanism;

[0013] The control modeling converts the decision results into specific control instructions and feeds them back to the perception module to form a closed loop.

[0014] Preferably, the perception module includes a visual submodule and an auditory submodule.

[0015] Preferably, the knowledge base includes a declarative knowledge table, a condition table and a result table, represents procedural knowledge in the form of "IF condition number THEN result number", and supports the association between conditions and results.

[0016] Preferably, the knowledge base is constructed by extracting rules from task operation procedures, the rule format is "if...then...", and the condition part and the result part are stored in a condition table and a result table respectively.

[0017] The method for calculating the cognitive load of personnel based on ACT-R includes the above-mentioned method for modeling the cognitive behavior of personnel based on ACT-R, and the calculation method further includes:

[0018] Based on the relationship between the time and weight of module activation in the ACT-R cognitive system, the instantaneous cognitive load of each module during task execution is calculated;

[0019] Based on the Logistic function, the error weight is dynamically calculated, which can reasonably describe the changes in cognitive load after people realize that an error has occurred;

[0020] Combining the SRK model and the back-propagation mechanism, the cognitive load weights of each module in the ACT-R cognitive model are determined;

[0021] During model operation, the instantaneous cognitive load of each module is dynamically calculated according to the task flow and the activation time of the knowledge base calling module;

[0022] The total cognitive load of the personnel is obtained by integrating and summing the instantaneous cognitive load of each module during the task execution, and the total cognitive load of the personnel includes the cumulative workload, the average workload and the long-term workload;

[0023] By using weight allocation and error correction mechanism, quantitative evaluation of the cognitive load of the personnel with different task complexity and personnel status is achieved.

[0024] Preferably, the calculation expression of the instantaneous cognitive load during the task execution is:

[0025] T(i,dt)=ψ i (t)×dt

[0026] where T(i,dt) is the instantaneous activation time function on module i, ψ i (t) is a function that determines whether module i is activated at a given time t, where 0 and 1 represent module inactivation and module activation, respectively;

[0027] Modify the Logistic function and derive the error weight function. The calculation expression is:

[0028]

[0029] Where L is the threshold, k is the steepness of the curve (the larger the k, the faster the growth rate), and t0 is the moment when the person realizes the error.

[0030] Preferably, the calculation expression of the instantaneous cognitive load of the module is:

[0031]

[0032] Where W total is the total instantaneous workload in time t, ω i is the weight assigned to module i, E i (t) is the error weight function. When no error occurs or the personnel are not aware of the error, the function value is 1. When the personnel are aware of the error, the error weight at the current moment is dynamically calculated.

[0033] Preferably, the cumulative workload represents the total workload generated by the operator during the execution of the task, which is calculated by integrating the instantaneous workload over time [0, T], where T is the total duration of the task, and the calculation process is as follows:

[0034]

[0035] Weight total It represents the sum of cognitive loads generated by all brain regions during the period of 0-T.

[0036] Preferably, the average workload is equal to the cumulative workload per unit time, and the calculation formula is as follows:

[0037]

[0038] Weight ave It represents the average workload. The module weights are allocated according to the SRK model, with the perception module weight being 1, the central module weight being 2, and the memory module weight being 4. The weight allocation is determined by verification based on the back-propagation mechanism to describe the cognitive load generated by each module per unit time.

[0039] Preferably, the calculation expression of the long-term load is:

[0040]

[0041] Where i represents each functional module, ω i is the cognitive load weight of module i, j is the activation duration of each module, E i (t i,j ) represents the error weight function of functional module i during the period j, and k represents the total working time of the module;

[0042] represents the additional cognitive load caused by long-term work on module i, and is calculated as follows:

[0043]

[0044] In the formula, n ensures that the function increases slowly in the first half, C represents the impact of fatigue state on personnel, the larger the C, the smaller the impact of fatigue state on personnel, k controls the overall inclination of the curve, the larger the k, the faster the cognitive load increases after personnel enter the fatigue state, x0 represents the time when personnel enter the fatigue state, the larger the x0, the slower the personnel enter the fatigue state, and the above variables are set to different values ​​according to personnel characteristics.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention is based on the ACT-R (Adaptive Control of Thought–Rational) cognitive system and derives a formula for calculating cognitive load. The nuclear power plant operating procedures are divided into declarative knowledge and procedural knowledge, and are entered into a knowledge base as a basis for the cognitive model to make decisions. When the cognitive model performs a task, the external environment acquired by the perception module is stored in a buffer, and then a knowledge search is performed. When the one that best matches the current environment is retrieved, the knowledge is used, and the action module is called to perform the corresponding operation. The cognitive load calculation formula embedded in the cognitive model can calculate the cognitive load generated when each module is called. The present invention provides a reference for further analyzing personnel cognitive behavior and improving work performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a block diagram of the operating mechanism of the ACT-R cognitive model of the present invention;

[0048] Figure 2 is an error weight curve diagram of the present invention;

[0049] Figure 3 Additional load function curve for long-term work

[0050] Figure 4 A block diagram describing the cognitive process of nuclear power plant accident diagnosis tasks of the present invention;

[0051] Figure 5 It is a block diagram of the cognitive behavior model of nuclear power plant accident diagnosis task of the present invention;

[0052] Figure 6 This is a task flow chart of the initial guidance part of the accident guideline in the first embodiment of the present invention;

[0053] Figure 7 The working principle of the cognitive system of the present invention and the overall structural block diagram of the reactor simulation system;

[0054] Figure 8 This is a flow chart of the nuclear power plant task guideline in the second embodiment of the present invention;

[0055] Fig. 9 This is a diagram of the operation steps of the primary circuit water capacity degradation cognitive model in the fourth embodiment of the present invention;

[0056] Fig.10 This is a diagram showing the activation duration of each module for the primary circuit water loading reduction in the fourth embodiment of the present invention;

[0057] Fig.11 This is a cognitive load diagram of each module for primary circuit water loading degradation in the fourth embodiment of the present invention;

[0058] Fig.12This is a diagram of the operation steps of the SGTR accident recognition model in the fourth embodiment of the present invention;

[0059] Fig.13 This is a diagram showing the activation duration of each module of an SGTR accident in the fourth embodiment of the present invention;

[0060] Fig.14 This is a cognitive load diagram of each module of the SGTR accident in the fourth embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Embodiment 1:

[0063] See also Figure 1 As shown in the figure, the personnel cognitive behavior modeling method based on ACT-R includes:

[0064] The cognitive process of personnel performing tasks is divided into three parts: perception modeling, decision modeling and control modeling;

[0065] The perception modeling obtains information from the external environment through the perception module, stores the information in the buffer, and processes it by the knowledge module;

[0066] The decision modeling activates production rules based on the information in the buffer, matches the corresponding knowledge base content, and selects applicable rules through a conflict resolution mechanism;

[0067] The control modeling converts the decision results into specific control instructions and feeds them back to the control module. The control module executes the instructions and changes the external environment. The perception module obtains information from the external environment to form a closed loop.

[0068] Specifically, the perception module includes a visual submodule and an auditory submodule, which are used to extract visual information from the display and auditory information from the alarm broadcast.

[0069] Specifically, the knowledge base includes a declarative knowledge table, a condition table and a result table, represents procedural knowledge in the form of "IF condition number THEN result number", and supports the association between conditions and results.

[0070] Specifically, the knowledge base is constructed by extracting rules from task operation procedures. The rule format is "if...then...", and the condition part and the result part are stored in a condition table and a result table respectively.

[0071] like Figure 4 and Figure 5 As shown in the figure, "perception" depends on the visual module and the auditory module, which extract valuable visual information from the display and alarm information from the broadcast respectively. The extracted information will be sent to the corresponding buffer, and the content in the buffer can be directly perceived by the cognitive model; the storage and processing of information depends on the declarative knowledge module and the procedural knowledge (procedural knowledge) module. "Decision-making" will query the declarative knowledge that matches the content of the target buffer to activate the production. The unmatched information will be discarded. If multiple production rules are matched, the conflict resolution mechanism will be triggered (since the operating procedures are in the form of processes, if a conflict occurs, the production rule closest to the last executed production rule will be selected); "control" is based on the output of the decision-making system and is ultimately reflected in the operator's operation of the reactor. From the input of information to the storage, processing and judgment of information, and finally the output of the decision result, the above cognitive process is constantly cyclical.

[0072] Since all the behaviors of the model are based on the knowledge base, the construction of the knowledge base is particularly important. By extracting knowledge from the nuclear power plant operating procedures, we can obtain knowledge in the form of "if...then...". 230 pieces of such knowledge were extracted from the accident diagnosis task. Taking the first six pieces as an example, some flow charts are shown below. Figure 6 As shown in the figure, if the condition is met, it goes to the node below, and if the condition is not met, it goes to the side node. The declarative knowledge number is marked on the upper right corner of the node.

[0073] 1) If the ELA / ELB loss alarm occurs, the EOR primary circuit is closed in the AC state and ELA+ELB is lost;

[0074] 2) If the ELA / ELB loss alarm does not appear, check whether the RT or RT signal appears;

[0075] 3) If RT or RT signal appears, check whether at least one main pump is running;

[0076] 4) If the RT or RT signal does not appear, check whether at least one SGS or MSS valve room temperature high 2 alarm appears;

[0077] 5) If at least one main pump is running, check whether ΔTsat (temperature saturation margin at the core outlet) is higher than ε;

[0078] 6) If no main pump is running, check whether ΔTsat is higher than -ε.

[0079] In order to facilitate knowledge management, reduce knowledge coupling, and facilitate cognitive model decision-making, the conditional part and the result part of the knowledge are split and stored separately. The storage forms of the above 6 pieces of knowledge in the database are shown in Tables 1, 2, and 3:

[0080] Table 1: Declarative knowledge table

[0081]

[0082] The declarative knowledge table stores all conditions and results. Declarative knowledge is operated by numbering, and it can be conveniently used as a condition or result.

[0083] Table 2: Condition table

[0084]

[0085] Table 3: Results table

[0086]

[0087] After constructing the condition table and result table, you can use the form of "IF condition number THEN result number" to express procedural knowledge, such as "IF T2 THEN J2". IF T2 means to remove the condition numbered T2 in the condition table. In the condition table, you can find its declarative knowledge K1, prefixed with NOT, which means "IF ELA / ELB lost the alarm and did not appear", and the THEN J2 part goes to the result table to query the record with the result numbered J2. It can be found that the declarative knowledge id corresponding to the result numbered J2 is K2, which means "THEN check whether RT or RT signal appears". Connecting the two in series is "IF ELA / ELB lost the alarm and did not appear THEN check whether RT or RT signal appears". This method can be used to express arbitrary procedural knowledge, thereby constructing the entire knowledge base.

[0088] According to the above method, the six pieces of knowledge can be expressed as:

[0089] 1) If the ELA / ELB loss alarm occurs, it will lead to the EOR primary circuit closed in the state AC condition loss of ELA+ELB → IF T1 THEN J1

[0090] 2) If the ELA / ELB loss alarm does not appear, check whether the RT or RT signal appears;

[0091] →IF T2 THEN J2

[0092] 3) If RT or RT signal appears, check whether at least one main pump is running;

[0093] →IF T3 THEN J3

[0094] 4) If RT or RT signal does not appear, check whether at least one SGS or MSS valve room temperature high 2 alarm appears; →IF T4 THEN J4

[0095] 5) If at least one main pump is running, check whether ΔTsat (temperature saturation margin at the core outlet) is higher than ε; →IF T5 THEN J5

[0096] 6) If no main pump is running, check whether ΔTsat is higher than -ε.

[0097] →IF T6 THEN J6.

[0098] Embodiment 2:

[0099] The method for calculating the cognitive load of personnel based on ACT-R includes the above-mentioned method for modeling the cognitive behavior of personnel based on ACT-R, and the calculation method further includes:

[0100] S1. Calculate the instantaneous cognitive load during task execution based on the relationship between the time and weight of module activation in the ACT-R cognitive system;

[0101] S2, dynamically calculate error weights based on the Logistic function;

[0102] S3, combining the SRK model and the back-propagation mechanism, the cognitive load weights of each module in the ACT-R cognitive model were determined;

[0103] S4. During the model operation, the instantaneous cognitive load of each module is dynamically calculated according to the task flow and the activation time of the knowledge base calling module;

[0104] S5. Integrate the instantaneous cognitive load during task execution and the instantaneous cognitive load of each module to obtain the total cognitive load of the personnel, where the total cognitive load includes the accumulated workload and the long-term workload;

[0105] S6. Use weight allocation and error correction mechanism to achieve quantitative evaluation of the cognitive load of the personnel with different task complexities and personnel status.

[0106] Specifically, the calculation expression of the instantaneous cognitive load during the task execution is:

[0107] T(i,dt)=ψ i (t)×dt

[0108] Where T (i,dt) is the instantaneous activation time function on module i, ψ i(t) is a function that determines whether module i is activated at a given time t, where 0 and 1 represent module inactivation and module activation, respectively;

[0109] like Figure 2 As shown in the figure, when an error occurs during task execution, the error weight function is activated and the error weight at each moment is dynamically calculated. The calculation method is as follows:

[0110]

[0111] In the formula, L represents the threshold, k represents the steepness of the curve (the larger k is, the faster the growth rate is), and t0 represents the moment when the operator realizes the error. The values ​​of L and k are set differently according to the characteristics of the personnel. For example, for employees with rich work experience, the k value can be set smaller and the L value can be set larger. According to experimental verification, L is 0.5-1.5 and k is 0.02-0.1, which is more in line with the general situation of operators.

[0112] Specifically, the calculation expression of the instantaneous cognitive load of the module is:

[0113]

[0114] Where W total is the total instantaneous workload in time t, ω i is the weight assigned to module i, E i (t) is the error weight function. When people make mistakes, they often cannot find them immediately, and they cannot solve the mistakes within a dt. It takes some time to solve them. In the process of solving the mistakes, the cognitive load of people should be high. represents the additional cognitive load caused by the error during this period, a represents the moment of realizing the error, and b represents the moment of solving the error.

[0115] Specifically, the cumulative workload represents the total workload generated by the operator during the execution of the task, which is calculated by integrating the instantaneous workload over time [0, T], where T is the total duration of the task, and the calculation process is as follows:

[0116]

[0117] Weight total It represents the sum of cognitive loads generated by all brain regions during the period of 0-T.

[0118] Specifically, the average workload is equal to the cumulative workload per unit time, and the calculation formula is as follows:

[0119]

[0120] Weightave Represents the average workload. In the ACT-R cognitive system, the cognitive cycle is usually set to 50ms. The duration of each module call during the model operation is a cognitive cycle. According to the above research on cognitive load weights, the weights of the eight modules of the cognitive module can be set. The visual module, auditory module, language module and motor module are used as motion perception modules, and their cognitive load weights are set to 1; the imagination module and the target module are used as central modules, and their cognitive load weights are set to 2; the declarative knowledge module and the procedural knowledge module are used as memory modules, and their cognitive load weights are set to 4.

[0121] like Figure 3 As shown, specifically, the calculation expression of the long-term load is:

[0122]

[0123] Where i represents each functional module, ω i is the cognitive load weight of module i, j is the activation duration of each module, E i (t i,j ) represents the error weight function of functional module i during the period of time j, k represents the total working time of the module, represents the additional cognitive load caused by long-term work on module i, and is calculated as follows:

[0124]

[0125] In the formula, n ensures that the function increases slowly in the first half, C represents the impact of fatigue on personnel, the larger the C, the smaller the impact of fatigue on personnel, k controls the overall inclination of the curve, the larger the k, the faster the cognitive load increases after the personnel enter the fatigue state, x0 represents the time when the personnel enter the fatigue state, the larger the x0, the slower the personnel enter the fatigue state, and the above variables are set to different values ​​according to the characteristics of the personnel. After experimental verification, n is generally set to 0.3-0.6, C is generally set to 750-1500, k is generally set to 0.002-0.01, and x0 is generally set between 1500-4000.

[0126] The cognitive load calculation formula requires the activation time of each module. Whether the calculation result is objective and accurate depends on whether the module activation is reasonable. A field is added to the declarative knowledge table to store the cognitive modules required for each operation. The corresponding modules are represented by the first letters of the words, visual module (V), auditory module (A), language module (S), action module (M), imagination module (I), goal module (G), declarative knowledge module (D), and procedural knowledge module (P). The modified declarative knowledge table is shown in Table 4:

[0127] Table 4: Modified declarative knowledge table

[0128]

[0129] To allocate each module, each step needs to be broken down into the most basic operation units. The following is the basis for allocating modules in the table above:

[0130] ELA / ELB loss alarm occurs: the operator is required to hear the alarm (auditory module) and determine that the alarm means ELA / ELB loss (declarative knowledge module).

[0131] RT or RT signal appears: The operator sees the signal appear (visual module) and determines that this signal represents RT or RT demand signal (declarative knowledge module).

[0132] At least one main pump is running: the operator sees a number (visual module), determines that this number represents the number of main pumps in operation (declarative knowledge module), and determines whether this number exceeds 1 (imagination module).

[0133] Guided to the EOR primary circuit closed in state AC condition and loss of ELA+ELB: the operator's accident diagnosis is completed, the results (declarative knowledge module) are given, and the operator is guided to specific accident guidelines.

[0134] Check whether RT or RT signal appears: The operator sets the task goal to check whether the RT signal appears (target module), uses the human-computer interaction component to open the corresponding data interface (action module), checks the signal (visual module), and determines whether this signal indicates the presence of the RT signal (declarative knowledge module)

[0135] Check whether at least one main pump is running: The operator sets the task goal to check whether the number of running main pumps is greater than 1 (target module), uses the human-computer interaction component to open the corresponding data interface (action module), checks the parameters (visual module), determines whether this parameter represents the number of running main pumps (declarative knowledge module), and determines whether this parameter exceeds 1 (imagination module).

[0136] Check whether at least one SGS or MSS valve room temperature high 2 alarm occurs: The operator sets the task goal to check whether at least one temperature high 2 alarm occurs (target module), uses the human-computer interaction component to open the corresponding data interface (action module), checks the signal (visual module), determines whether this signal represents a temperature high 2 alarm signal (declarative knowledge module), and determines whether the number of alarms exceeds 1 (imagination module).

[0137] Check whether ΔTsat (temperature saturation margin at the core outlet) is higher than ε: The operator sets the task goal to check whether ΔTsat is higher than ε (target module), uses the human-computer interaction component to open the corresponding data interface (action module), checks the parameters (visual module), determines whether this parameter represents ΔTsat (declarative knowledge module), and determines whether ΔTsat exceeds ε (imagination module).

[0138] Check whether ΔTsat is higher than -ε: The operator sets the task goal to check whether ΔTsat is higher than -ε (target module), uses the human-computer interaction component to open the corresponding data interface (action module), checks the parameters (visual module), determines whether this parameter represents ΔTsat (declarative knowledge module), and determines whether ΔTsat exceeds -ε (imagination module).

[0139] When each procedural knowledge (production rule) is activated, the procedural knowledge module needs to be called once after the condition is executed. Each time a cognitive module is called, it means that the module has activated a cognitive cycle. In order to simulate the real situation, a random disturbance of plus or minus 50ms is added to it.

[0140] The reactor simulation system is independent of the cognitive modeling simulation system and uses different frameworks and communication protocols. The native reactor simulation system does not provide a cognitive model access interface, so the adapter mode is used to coordinate the communication between the two parties. The working principle of the cognitive system and the overall structure of the reactor simulation system are as follows: Figure 7 shown.

[0141] like Figure 7 As shown in the figure, the interaction mode between the ACT-R cognitive model and the reactor simulation system is mainly introduced. The "external environment" on the left represents the reactor simulation system; the "communication adapter module" in the middle is responsible for reading the data of the reactor simulation system in real time and sending it to the ACT-R cognitive model; the "cognitive system module" on the right represents the ACT-R cognitive model, which interacts with the reactor system through the perception module and the motion module, simulating a real operator using human-computer interaction components (such as display, mouse, keyboard, etc.) to operate the reactor system. The decision of the cognitive model is based on the knowledge base.

[0142] Since the task guideline process of nuclear power plants is often very complex and the amount of procedural knowledge extracted is huge, the modeling process becomes extremely cumbersome and too professional. Only developers can build cognitive models, and they are not easy to maintain. To address this problem, our team built a cognitive modeling platform website. The website is implemented using Django+Vue, so that users without programming knowledge can also build cognitive models. The extracted knowledge is stored on the server for easy subsequent calls to the model. The model operation platform is mainly based on the ACT-R cognitive system kernel, developed in Python, and the platform interface is implemented using PyQt. The cognitive model communicates with the ACPR50S experimental reactor simulation platform through the WebSocket protocol. When the model is running, a Socket request is sent to the simulation reactor platform to obtain reactor data, and then a request is sent to the Django backend of the cognitive modeling platform to obtain data in json format, and parse it into procedural knowledge that the model can call, and decide the next operation based on the production rules. The overall structure is as follows: Figure 8 As shown:

[0143] Figure 8 Description: This paper introduces the relationship between the reactor simulation platform, ACT-R cognitive model, modeling platform, and database. The operation of the ACT-R cognitive model is based on the knowledge base, so it is necessary to send a request to the modeling platform to obtain the knowledge base content. The knowledge base of the modeling platform is stored in the database.

[0144] From the above, it can be seen that the present invention is based on the ACT-R (Adaptive Control of Thought–Rational) cognitive system, and derives the cognitive load calculation formula; the nuclear power plant operating procedures are divided into declarative knowledge and procedural knowledge, and entered into the knowledge base as the basis for the cognitive model to make decisions. When the cognitive model performs a task, the external environment obtained by the perception module is stored in the buffer, and then a knowledge search is performed. When the one that best matches the current environment is retrieved, the knowledge is used, and the action module is called to perform the corresponding operation. The cognitive load calculation formula embedded in the cognitive model can calculate the cognitive load generated when each module is called. The present invention provides a reference for further analyzing personnel cognitive behavior and improving work performance.

[0145] Embodiment three:

[0146] Experimental design:

[0147] A total of 4 subjects were recruited, aged 27-42 years old, all of whom were nuclear power plant operators with rich work experience. The subjects performed accident diagnosis tasks on the ACPR50S experimental reactor simulation platform. Before the formal experiment, it was confirmed that the subjects could master the operation methods of the simulation platform proficiently. Each subject will conduct two experiments, with an interval of more than 1 hour between the two experiments to ensure that the subjects have sufficient rest. The content of one experiment is: the experimenter sets the accident on the experimental reactor simulation platform, and the subject infers the current accident through the operating parameters. During the task, it is necessary to record the operation and timestamp of each step of the subject, so as to compare it with the operation steps when the cognitive model performs the same task; after the experiment, fill in the NASA-TLX scale immediately. Valid data is based on the correct operation of the subjects, and the erroneous operation data will not be counted in the valid data range. The cognitive load results calculated by the NASA-TLX scale are used as the benchmark, and the correlation analysis is performed with the calculation results of the cognitive model. Two tasks were selected in this experiment, namely the first circuit load degradation accident and the rupture of a heat transfer tube (SGTR) accident.

[0148] Model validity analysis:

[0149] A typical accident was selected for the experiment, and the effectiveness of the model was examined by comparing the consistency between the operating steps and diagnostic results of the subjects and the model when facing the same accident. Tables 5 and 6 respectively record the operating steps, parameter states, and diagnostic results of the operator and the cognitive model when performing the initial guidance task of the accident guideline under the condition of primary water level degradation.

[0150] Table 5: Test results - initial guidance of accident guidelines (primary circuit water loading reduction)

[0151]

[0152] Table 6: Cognitive model operation results - initial guidance of accident guidelines (primary circuit water capacity reduction)

[0153]

[0154]

[0155] Tables 7 and 8 respectively record the operation steps, parameter states, and diagnostic results of the operator and the cognitive model when performing the initial guidance task of the accident guidelines when a spiral tube ruptures (SGTR accident) and the steam generator is radioactive.

[0156] Table 7: Test results - initial guidance of accident guidelines (SGTR accident)

[0157]

[0158]

[0159] Table 8: Cognitive model operation results - accident guideline initial guidance (SGTR accident)

[0160]

[0161]

[0162]

[0163] By comparing Table 5 with Table 6, and Table 7 with Table 8, it can be found that the operation steps and diagnosis results of the cognitive model and the operator are consistent, which can prove the effectiveness of the cognitive model.

[0164] Embodiment 4:

[0165] Analysis of the effectiveness of cognitive load formula:

[0166] The most mainstream cognitive load evaluation method in the world, NATA-TLX scale, is used as a comparison benchmark. The cognitive load calculation results of the cognitive model are compared with those of the NASA-TLX scale to prove the validity of the cognitive load formula. The subjects compared the six dimensions of the NASA-TLX scale in pairs and selected the dimensions that they thought were more closely related to cognitive load. The weights of each dimension were calculated as follows: mental demand 17.3%, physical burden 9.3%, time demand 10.7%, task performance 28%, effort 18.7%, and frustration 16%. The final score of the NASA-TLX scale can be obtained by multiplying the weights and the scores of each item. As shown in Table 9, the scale score represents the cognitive load level of the operator.

[0167] Table 9: Statistical results of operator cognitive load under two accident conditions

[0168]

[0169]

[0170] The reliability and validity of the scale were analyzed using SPSS data analysis software. The reliability α was 0.920, which was in the excellent range; the validity KMO was 0.708, which had good validity.

[0171] When the accident is set as the degradation of the primary circuit water loading on the ACPR50S experimental reactor simulation platform, the operation steps of the cognitive model are as follows: Fig. 9 As shown in Figure 2, the cognitive load calculation results are as follows: Fig.10 , 11 As shown;

[0172] The calculation results of the mental workload of the cognitive model in processing the degradation of the primary water supply are shown in Table 10. The total activation time of each cognitive module in processing the primary water supply degradation accident is 1299.857ms, and the total cognitive load generated is 14.031. The experiment was repeated three times, and the calculated cognitive load results were 14.819, 12.394, and 15.094 respectively.

[0173] Table 10: Calculation results of mental workload in primary circuit water capacity degradation accidents

[0174]

[0175] When the accident is set as a single heat transfer tube rupture (SGTR) accident on the ACPR50S experimental reactor simulation platform, the cognitive model operation steps are as follows: Fig.12 As shown in Figure 2, the cognitive load calculation results are as follows: Fig.13 , 14 As shown;

[0176] The calculation results of the mental workload of the cognitive model in processing SGTR accidents are shown in Table 11. The total activation time of each cognitive module in processing SGTR accidents is 2800.012ms, and the total cognitive load generated is 44.773. The experiment was repeated three times, and the calculated cognitive load results were 48.657, 45.495, and 47.488 respectively.

[0177] Table 11: Calculation results of mental workload in SGTR accidents

[0178]

[0179] The overall operation screenshot of the simulation platform is as follows: Fig.13 As shown;

[0180] Result analysis:

[0181] A correlation analysis was performed between the cognitive load calculated by the NASA-TLX scale and the cognitive load calculated by the cognitive model. The analysis results are shown in Table 12.

[0182] Table 12: Correlation analysis between NASA-TLX scale cognitive load and cognitive model calculation results

[0183]

[0184] The correlation analysis result Sig is 0.012, which is less than 0.05, indicating that there is a significant correlation between the two, and the correlation coefficient is 0.822. This shows that the cognitive load calculated by the NASA-TLX scale is significantly correlated with the calculation results of the cognitive model, and is positively correlated. Therefore, the cognitive compliance calculation formula can effectively evaluate the operator's cognitive load.

[0185] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0186] In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other.

[0187] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The personnel cognitive behavior modeling method based on ACT-R is characterized by: The following steps are involved: The cognitive process of personnel performing tasks is divided into three parts: perception modeling, decision modeling and control modeling; The perception modeling obtains information from the external environment through the perception module, stores the information in the buffer, and processes it by the knowledge module; The decision modeling activates production rules based on the information in the buffer, matches the corresponding knowledge base content, and selects applicable rules through a conflict resolution mechanism; The control modeling converts the decision results into specific control instructions and feeds them back to the perception module to form a closed loop.

2. The personnel cognitive behavior modeling method based on ACT-R according to claim 1 is characterized in that: The perception module includes a visual submodule and an auditory submodule.

3. The personnel cognitive behavior modeling method based on ACT-R according to claim 1, characterized in that: The knowledge base includes a declarative knowledge table, a condition table and a result table, represents procedural knowledge in the form of "IF condition number THEN result number", and supports the association between conditions and results.

4. The personnel cognitive behavior modeling method based on ACT-R according to claim 1 is characterized in that: The construction of the knowledge base is generated by extracting rules from the task operation procedures. The rule format is "if...then...", and the condition part and the result part are stored in the condition table and the result table respectively.

5. The method for calculating personnel cognitive load based on ACT-R is characterized by: The method comprises the personnel cognitive behavior modeling method based on ACT-R as claimed in any one of claims 1 to 4, wherein the calculation method further comprises: Based on the relationship between the time and weight of module activation in the ACT-R cognitive system, the instantaneous cognitive load of each module during task execution is calculated; Based on the Logistic function, the error weight is dynamically calculated, which can reasonably describe the changes in cognitive load after people realize that an error has occurred; Combining the SRK model and the back-propagation mechanism, the cognitive load weights of each module in the ACT-R cognitive model are determined; During model operation, the instantaneous cognitive load of each module is dynamically calculated according to the task flow and the activation time of the knowledge base calling module; The total cognitive load of the personnel is obtained by integrating and summing the instantaneous cognitive load of each module during the task execution, and the total cognitive load of the personnel includes the cumulative workload, the average workload and the long-term workload; By using weight allocation and error correction mechanism, quantitative evaluation of the cognitive load of the personnel with different task complexity and personnel status is achieved.

6. The method for calculating cognitive load of personnel based on ACT-R according to claim 5, characterized in that: The calculation expression of the instantaneous cognitive load during the task execution is: T(i,dt)=ψ i (t)×dt where T(i,dt) is the instantaneous activation time function on module i, ψ i (t) is a function that determines whether module i is activated at a given time t, where 0 and 1 represent module inactivation and module activation, respectively; Modify the Logistic function and derive the error weight function. The calculation expression is: Where L represents the threshold, k represents the steepness of the curve, the larger the k, the faster the growth rate, and t0 represents the moment when the person realizes the error.

7. The method for calculating cognitive load of personnel based on ACT-R according to claim 5, characterized in that: The calculation expression of the instantaneous cognitive load of the module is: Where W total is the total instantaneous workload in time t, ω i is the weight assigned to module i, E i (t) is the error weight function. When no error occurs or the person is not aware of the error, the function value is 1. When the person is aware of the error, the error weight at the current moment is dynamically calculated.

8. The method for calculating cognitive load of personnel based on ACT-R according to claim 5, characterized in that: The cumulative workload represents the total workload generated by the operator during the mission, and is calculated by integrating the instantaneous workload over time [0, T], where T is the total mission duration, as follows: Weight total It represents the sum of cognitive loads generated by all brain regions during the period of 0-T.

9. The method for calculating cognitive load of personnel based on ACT-R according to claim 5, characterized in that: The average workload is equal to the cumulative workload per unit time, and is calculated as follows: Weight ave It represents the average workload. The module weights are allocated according to the SRK model, with the perception module weight being 1, the central module weight being 2, and the memory module weight being 4. The weight allocation is determined by verification based on the back-propagation mechanism to describe the cognitive load generated by each module per unit time.

10. The method for calculating cognitive load of personnel based on ACT-R according to claim 5, characterized in that: The calculation expression of the long-term load is: Where i represents each functional module, ωi represents the cognitive load weight of module i, j represents the activation time of each module, E i (t i,j ) represents the error weight function of functional module i during the period j, and k represents the total working time of the module; represents the additional cognitive load caused by long-term work on module i, and is calculated as follows: In the formula, n ensures that the function increases slowly in the first half, C represents the impact of fatigue state on personnel, the larger the C, the smaller the impact of fatigue state on personnel, k controls the overall inclination of the curve, the larger the k, the faster the cognitive load increases after personnel enter the fatigue state, x0 represents the time when personnel enter the fatigue state, the larger the x0, the slower the personnel enter the fatigue state, and the above variables are set to different values ​​according to personnel characteristics.

Images