Improved CREAM-based dynamic and static collaborative evaluation method for human reliability of underwater aircraft

Abstract

The invention provides an improved CREAM-based dynamic and static collaborative evaluation method for the human reliability of an underwater vehicle, and the method comprises the following steps: determining a scene environment performance condition CSEPC based on a constructed scene environment risk factor pedigree, obtaining a cognitive function error risk influence coefficient through a decision laboratory method, determining a cognitive function risk judgment factor, and carrying out the judgment of the cognitive function error risk. Calculating a scene risk comprehensive index; constructing an improved CREAM risk prediction model according to the scene risk comprehensive index S, obtaining a cognitive error probability, and introducing adaptive fuzzy logic to correct the error probability; based on the cognitive error probability, multi-source observation information is introduced for fusion, a dynamic Bayesian network is used for real-time updating, and a final cognitive error probability of a task time level is obtained in combination with a dynamic environment performance correction factor. According to the method, the dependence on static general performance conditions is reduced, and the evaluation stability can be kept under the condition of data noise, missing or conflict.

Description

Technical Field

[0001] This invention belongs to the field of reliability analysis technology, and in particular relates to a dynamic and static collaborative evaluation method for the human factors reliability of submariners based on an improved CREAM. Background Technology

[0002] Human factors play a significant role in accidents occurring in complex human-machine systems. Human error is considered a major cause of accidents and disasters, occurring in various industries from nuclear accidents and aerospace engineering to space missions and medicine. Human error is one of the primary causes of risk events. Manned submersibles are complex human-machine systems with intricate human-machine collaborations involving multiple systems such as ship, machinery, control, electrical, and underwater acoustic communication. The deep-sea operating environment also places higher demands on the safety of submersible crews. Deep-sea manned submersible underwater operations are complex systems engineering projects, characterized by complex human-machine relationships, harsh operating environments, and long operating times. Therefore, introducing human factor reliability theory and methods to explore the impact of extreme deep-sea environmental factors on personnel operational capabilities is particularly important. Human Reliability Analysis (HRA), as a key method for assessing and reducing the risk of human error, has been widely applied in high-risk industries such as nuclear power, aerospace, and maritime. However, traditional HRA methods are difficult to adapt to the complex environment of the deep sea. Therefore, how to build a more accurate human error prediction model for manned submersibles, tailored to the special environmental characteristics of manned submersibles, has become a technical problem that urgently needs to be solved.

[0003] Cognitive Reliability and Error Analysis (CREAM) is a typical representative of second-generation human factor reliability analysis methods. Its core idea is that human performance ability is not an isolated random behavior but depends on the context in which the task is performed. It ultimately determines human response behavior by influencing cognitive control patterns and their effects across different cognitive activities. CREAM provides two methods: the basic method and the extended method. The basic method is used to determine control patterns and corresponding error rate intervals by screening basic levels of human factor errors, while the extended method can be used to quantify cognitive function errors. However, existing technologies have the following problems:

[0004] (1) Limitations of the cognitive model. The CREAM method uses a cognitive model of "observation, interpretation, judgment, and execution," which cannot fully describe the human cognitive process. For example, in actual work, operators acquire stimulus signals in many ways, and observation is one of the main means of information acquisition, which can only obtain visual stimulus signals. This cognitive model lacks data on sound, touch, vibration, and other signal stimuli.

[0005] (2) Over-reliance on general performance conditions. CREAM's assessment of the scenario environment mainly relies on the nine indicators provided by the general performance conditions. Although the application of the nine indicators in the nuclear power industry has proven their effectiveness, when the CREAM method is used to assess human factors reliability in other fields, the nine indicators cannot provide a complete description of the scenario environment.

[0006] (3) It cannot assess the human factor reliability of dynamic processes. The CREAM method itself is biased towards static assessment, and its scenario-dependent control model is a static description, which cannot assess the human factor reliability of dynamic environments or dynamic task processes. If dynamic environment or task information is simply applied to this method, it is impossible to determine the impact of dynamic environment or task conditions on personnel's work capabilities. Summary of the Invention

[0007] The purpose of this invention is to address the limitations of existing cognitive models, their over-reliance on general performance conditions, their inability to assess the human factor reliability of dynamic processes, and their lack of consideration for human cognitive behavioral processes in task descriptions.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A dynamic and static collaborative assessment method for the human factors reliability of submariners based on an improved CREAM includes the following steps: S1. Based on the constructed scenario-environment risk factor spectrum, determine the scenario-environment performance condition CSEPC, use the decision laboratory method to obtain the cognitive function error risk impact coefficient k, and determine the cognitive function risk judgment factor. Then, the scenario risk comprehensive index S is calculated, and the risk control model and error probability range are obtained based on the scenario risk comprehensive index S. S2. Obtain the probability of cognitive error based on the comprehensive situational risk index S, and introduce adaptive fuzzy logic to correct the probability of error. S3. Based on the cognitive error probability obtained in step S2, multi-source observation information is introduced for fusion, and a dynamic Bayesian network (DBN) is used for real-time updates, combined with a dynamic environmental performance correction factor. This allows us to obtain the final cognitive error probability at the task moment level.

[0009] Furthermore, step S1 specifically includes: S101. Construct a spectrum of scenario-environment risk factors and use information entropy to calculate the weights of scenario performance condition risk factors in the spectrum of scenario-environment risk factors. The situational environment performance condition (CSEPC) was screened and determined by assessing the impact of four cognitive functions on risk factors in the risk factor spectrum of the situational environment, and the risk impact coefficient k of cognitive function failure was obtained by using the decision laboratory method. S102. Using fuzzy trigonometric functions to determine cognitive function risk assessment factors. ; S103. Based on the risk factor weights of scenario performance conditions The risk impact coefficient k of cognitive function failure and the comprehensive risk index S of the cognitive function risk assessment factor γ are obtained from the scenario risk: ; In the formula, It is the first A comprehensive index of situational risk for cognitive functions. It is the first Item scenario performance condition risk factor weights, It is the first Scenario, environment, and performance conditions for the first The impact coefficient of the risk of cognitive function failure. It is the first Cognitive function risk assessment factors for performance conditions in specific scenarios and environments; S104. Based on the scenario risk comprehensive index S, establish a correspondence model between the scenario environment performance conditions of manned submersibles and risk control modes, and determine the risk control mode and the error probability range.

[0010] Furthermore, step 101 specifically includes: S1011. Based on the characteristics of complex scenarios and environments for manned submersibles, the risk factors of the scenario and environment are processed and divided into five categories: organizational factors, task factors, personal factors, human-machine factors and environmental factors, and a spectrum of risk factors for complex scenarios and environments of manned submersibles is constructed. S1012. Using information entropy, calculate the weights ω of each situational performance condition risk factor for each situational performance condition risk factor and the information elements of the four core cognitive functions. ; In the formula This represents the total number of information elements in the system. For information elements Normalized weights; For information elements The entropy value, ; For information elements The characteristics; For each information element The probability of the state occurring, Indicates the possible states of an information element; Based on the weight ω of each scenario performance condition risk factor, the secondary indicators of the scenario performance condition risk factors are ranked to obtain the scenario environment performance condition CSEPC. S1013. The Delphi method is used to score the degree of influence between each pair of information elements, and the direct influence matrix is ​​obtained based on the scoring results. The normalized influence matrix M is obtained by performing normalization. ; The overall relationship matrix is ​​obtained by calculating the standardized influence matrix: ; In the formula It is the identity matrix; Overall relational matrix; In the decision laboratory method, centrality is used as the basis for decision-making. and causal degree As two indicators, the calculation formula is: ; In the formula: Indicates the degree of influence of element i; This indicates the degree to which element i is affected; It represents the centrality of element i in the system. The larger the value, the more it is related to other factors. This represents the tendency of element i to influence other elements; n is the sample size. Represents the overall relation matrix Some elements in; Expert evaluation was used to score the impact of the Scenario-Environment Performance Conditions (CSEPC) on the four cognitive functions of submariners, resulting in a score for the impact of each CSEPC condition on cognitive function. This allows for the determination of the impact coefficient of cognitive impairment risk. ; .

[0011] Furthermore, step 102 specifically involves: The risk level of each scenario-environment performance condition (CSEPC) is characterized within a fuzzy set. Each scenario-environment performance condition (CSEPC) has three types of impact on human reliability: increasing, neutral, and decreasing. Corresponding membership functions are then constructed accordingly. ; ; ; In the formula, To reduce risk, To be neutral on the impact of risk, To increase risk; Based on membership functions, obtain the weights of the impact of situational performance conditions (CSEPC) on human reliability: ; The cognitive function risk assessment factor is ultimately obtained based on the weights. : ; In the formula As a risk assessment factor for cognitive function, The average score given by the experts.

[0012] Furthermore, step S2 specifically involves: (1) Introduce adaptive fuzzy logic method, task complexity C and environmental change E fuzzy membership function into the scenario risk comprehensive index S. and Obtain the probability of cognitive error (CFP*); ; ; in, It is an adaptive fuzzy correction factor; To adjust the coefficient, This is the sensitivity coefficient of task complexity to error probability; This is the sensitivity coefficient of the probability of error to environmental changes; (2) Based on the weight values ​​of each cognitive function, and considering the weights of the probability of occurrence of each error pattern, obtain the final basic value of the error probability: ; In the formula The basic error probability of cognitive behavior; Weights for perceptual and cognitive functions; For cognitive function weights; Weights for decision-making cognitive functions; Weights reflecting cognitive function; Let be the weight of the i-th error pattern under cognitive function G; Let be the basic error probability of the i-th term under cognitive function G; Let J be the weight of the j-th error pattern under cognitive function R; Let be the basic error probability of the j-th term under cognitive function R; For the weight of the i-th error pattern under cognitive function J; Let J be the basic error probability of the j-th term under cognitive function J; The weight of the h-th error pattern under cognitive function F; Let h be the basic error probability of the h-th term under cognitive function F.

[0013] Furthermore, step S3 specifically includes: S301. Based on the DS evidence theory, trust assignment and basic probability assignment are fused to obtain the observations at time t. ; S302. Construct a dynamic Bayesian network DBN and update the probability of cognitive error in real time based on the dynamic Bayesian network DBN. S303. Based on the multi-parameter environmental suitability analysis method, a dynamic environmental performance correction factor was constructed. The derivation model is as follows: a normalization function is constructed through membership mapping and weighted fusion to correct the cognitive error probability output by DBN. S304. The final cognitive error probability is calculated by combining the dynamic environment performance correction factor and the cognitive error probability output by DBN.

[0014] Furthermore, step S301 specifically includes: (1) Convert single-source observations into identification frameworks Basic probability assignment (BPA); Suppose the identification frame is for the first... Standardized score of each information source at time t Given the trust or reliability weight of the information source ∈[0,1], its BPA is defined as: ; In the formula For the quality allocation of propositions by information source s at time t, satisfying ; To identify the framework, N represents normal and R represents risk; (2) The Dempster combination rule is used to analyze the two information sources. p-fusion: ; In the case of multiple information sources, the data is iteratively fused sequentially to obtain the total BPA after fusion: ; In the formula Let A, B, and C be the basic probability assignments of information sources o and p at time t; and let C be the recognition frames. A subset of; To determine the quality of proposition A after the fusion of the two sources; Let O be the basic probability assignment of the o-th information source to the proposition set B at time t; Let p be the basic probability assignment of the p-th information source to the proposition set C at time t; The degree of conflict between the evidence from the two sources; For all The total BPA after fusion of information sources at time t; (3) Map the fusion results to observations , The specific formula is as follows: ; In the formula For the quality of trust in risk propositions; The betting probability is obtained by distributing uncertainty equally between risk and normal. It is the probability of betting on the proposition "risk".

[0015] Furthermore, step S302 specifically includes: (1) The dynamic Bayesian network contains three time-rolling layers: latent variables ; Observation ; Errors; (2) Let Let be the comprehensive risk index for the scenario at time t. , ∈[0,1] represent the fuzzy membership degrees of task complexity and environmental change, respectively. Then: ; In the formula Let be the probability that the cognitive control mode transitions from the previous state u to the current state v at time t; u is the mode at the previous time; v is the candidate mode at the current time; w is all the candidate modes in the denominator, used for normalization. The transition probability is time-varying. For the intercept term, These are the scenario risk composite indexes. Task complexity and fuzzy membership Fuzzy membership degree due to environmental changes The coefficient; (3) Given When each observation dimension follows a Gaussian distribution and is conditionally independent, a conditional Gaussian observation model is obtained: ; In the formula Indicating in cognitive control mode Under known conditions, the observation The probability distribution; Control modes Next The mean and variance parameters of the dimensional features; (4) Given the time-varying transition probability and the conditional Gaussian observation model, the posterior distribution of the cognitive control mode is updated recursively using forward filtering; the recursive relationship is as follows: ; In the formula This is a cognitive control model; and It indicates a transition from one mode in the previous moment to another mode in the current moment; To predict priors; It is a forward-posterior distribution; The T-1 submariner is in control mode at this time. The posterior probability of ; (5) Conditional error rate of each mode Perform a full probability weighting; the weighted result is the probability of cognitive error of DBN at time t: ; Furthermore, step S303 specifically includes: Trapezoidal fuzzy membership functions are used to map the suitability of each dynamic cabin environment parameter to the [0,1] interval; for the th Parameters Its membership function is: ; In the formula: For the first Measured values ​​of dynamic cabin environmental parameters For a specific moment; For parameters The corresponding uniformized membership value ranges from [0,1]. , , , Based on the spectrum of environmental risk factors, as well as the general medical requirements for temperature environment in working compartments (GJB 898-90) and the permissible concentration of air components in conventional submarines (GJB11B-2012), these four key thresholds are set. Setting the first The weight of the item parameter is This weight is obtained through expert scoring and reflects the relative importance of each dynamic parameter to the overall cognitive performance of the submariner, and satisfies the following: ; Obtain the suitability index of dynamic cabin environmental parameters. , represented as: ; In the formula, N represents the total number of selected dynamic cabin environmental parameters; Let p be the weight of the p-th dynamic cabin environment parameter on the overall cognitive performance of the submariner; The suitability index of dynamic cabin environmental parameters Mapped to cognitive error correction factor : ; constant This determines the maximum amplification factor of the probability of cognitive error under the most unfavorable environmental conditions.

[0016] Furthermore, step S304 specifically includes: Based on cognitive error correction factor The final probability of cognitive error is obtained as follows: ; In the formula, This represents the probability of a final cognitive error at time t; This represents the probability of cognitive error at time t obtained by forward recursion of DBN; This is a dynamic environmental performance correction factor.

[0017] The beneficial effects of this invention are as follows: 0. Addressing the limitations of existing CREAM methods, which rely on a single cognitive model structure encompassing observation, interpretation, judgment, and execution, and fail to comprehensively reflect the multimodal cognitive processes of submariners, this invention introduces four cognitive functions: perception, cognition, decision-making, and reaction. It also incorporates situational environmental performance conditions to construct a comprehensive situational risk index. This model fully characterizes the cognitive behavioral processes of submariners in complex tasks, significantly improving the comprehensiveness and accuracy of cognitive modeling and providing a cognitive foundation for human factors reliability assessment that is closer to actual operational scenarios.

[0018] 1. To address the issue that the CREAM method relies excessively on nine general performance conditions and is ill-suited to the complex environment of the deep sea, this invention constructs a scenario-environment risk factor spectrum encompassing five primary indicators (organization, task, individual, human-machine, and environment) and 14 secondary indicators. It then combines information entropy and the E-DEMATEL method for weight calculation and factor selection. This method effectively reduces reliance on a single general performance condition, enhances the applicability and universality of the assessment system in the extreme environment of the deep sea, and makes the assessment results more targeted and reliable.

[0019] 2. To address the limitation of the CREAM method in effectively assessing dynamic task processes, this invention introduces dynamic Bayesian networks and DS evidence theory on top of static assessment, enabling real-time fusion and dynamic updating of multi-source observational information on environment, behavior, and physiology. By constructing a time-varying transition probability model and a conditional Gaussian observation model, combined with a forward filtering recursive mechanism, the stability and robustness of the assessment results can be maintained even with noise, missing data, or conflicts, significantly improving the real-time monitoring and early warning capabilities for human reliability during dynamic tasks. By fusing multi-source observational information through DS evidence theory and dynamically correcting for task complexity and environmental changes using adaptive fuzzy logic, a "dynamic-static collaborative" assessment system is constructed. This method not only retains the structured advantages of the CREAM method but also possesses capabilities such as online updating, environmental adaptation, and noise robustness, significantly improving the accuracy and practicality of assessing human reliability for submariners in complex deep-sea mission scenarios.

[0020] 3. To address the issue that traditional methods neglect the immediate impact of dynamic environments on human performance, this invention constructs a dynamic environmental performance correction factor α based on multi-parameter environmental suitability analysis. This factor maps key environmental parameters such as temperature, humidity, and gas concentration to cognitive error correction factors using a trapezoidal fuzzy membership function. This mechanism can respond in real-time to changes in the cabin environment, dynamically adjust the probability of cognitive errors, and significantly enhance the adaptability and early warning sensitivity of the assessment method to extreme or changing environments. Attached Figure Description

[0021] Figure 1 This is a flowchart of the static and dynamic collaborative evaluation method provided by the present invention; Figure 2 This is a diagram of the risk index prediction model based on the contextual cognitive behavior model provided by the present invention; Figure 3 This is a framework diagram of the Cognitive Behavioral Segregation (CSE) model provided by the present invention; Figure 4 This is a diagram of the improved CREAM risk prediction model provided by the present invention; Figure 5 This is a schematic diagram illustrating the relationship between the scenario risk comprehensive index and the risk control model provided by the present invention; Figure 6 It is the membership function of the fuzzy set of CSEPCs provided by this invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0023] The application principle of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0024] Please refer to Figures 1-4 A dynamic and static collaborative assessment method for the human factors reliability of submariners based on an improved CREAM framework includes: S1. A method for predicting human error among submarine crew members is applied to the prediction of scenario-based risks in operational systems operating under special and complex circumstances. Specifically, it involves: firstly, based on the risk factor weights ω of each scenario performance condition in the scenario-based risk factor spectrum, evaluating the impact of the four core cognitive functions of the cognitive behavior model on risk factors to screen and determine the scenario-based performance conditions (CSEPC); and then using E-DEMATEL to obtain the risk impact coefficient of cognitive function errors. And fuzzy trigonometric functions were used to determine the risk assessment factors for cognitive function. Based on this, a comprehensive scenario risk index S is obtained using three indicators: weight ω, the impact coefficient of cognitive function error risk k, and the cognitive function risk judgment factor γ. This ultimately determines the risk control model and provides the error probability range. The prediction model is as follows: Figure 5 As shown.

[0025] S101. Based on the risk factor weights ω of each situational performance condition in the situational environment risk factor spectrum, the situational environment performance condition CSEPC is screened and determined by assessing the impact of the four cognitive functions (perception, cognition, decision-making, and response) of the cognitive behavior model on the risk factors. The impact coefficient of cognitive function failure risk is then obtained using causal relationship analysis through the decision laboratory method (DEMATEL). .

[0026] S1011. In view of the complex scenario environment characteristics of manned submersibles, the scenario environment risk factors that may induce human error are systematically sorted out and classified into five categories: organizational factors, task factors, personal factors, human-machine factors and environmental factors. A spectrum of risk factors in complex scenario environments for manned submersibles is constructed, which includes 14 scenario performance condition risk factors and four core cognitive functions, as shown in Table 1.

[0027] Table 1. Spectrum of Scenario-Environmental Risk Factors

[0028] S1012. Using the E-DEMATEL method, namely the information entropy and decision laboratory method, the weights ω of each situational performance condition risk factor and the information elements of the four core cognitive functions are calculated by using information entropy to obtain the situational environment performance condition CSEPC.

[0029] (1) Obtain the risk factor weights ω of the scenario performance conditions by calculating the information element weights based on Entropy.

[0030] a. Computational Information Elements entropy value To reduce the bias of subjective scoring, the information entropy method is used to calculate the importance of each information element. Expert scoring is used to obtain the status of information elements for each scenario's performance conditions, risk factors, and four core cognitive functions (perception, cognition, decision-making, and response). Using the information entropy method, the uncertainty of each information element is quantified based on its discreteness; the lower the uncertainty, the greater the amount of information contained in that information element, and the higher its importance for the assessment. Each information element is defined... The probability of this state occurring is ,in Entropy value represents the possible states of an information element. Information element The entropy value is calculated using the following formula: (1); b. Normalized entropy value: The entropy value of each information element is normalized, and the weight ω of the risk factor for each scenario performance condition is calculated. (2); In the formula This represents the total number of information elements in the system. For information elements The normalized weights are used to adjust the importance of information elements in subsequent steps; For information elements The entropy value; For information elements Its characteristics.

[0031] (2) Based on the weight ω of each scenario performance condition risk factor, the secondary indicators of the 14 scenario performance condition risk factors are ranked and the important influencing factors are screened out. Ten key risk factors are obtained by ranking the scenario performance condition risk factors according to their weights, and they are renumbered C1-C10, which are the ten scenario environmental performance conditions CSEPC.

[0032] S1013. Obtain the influence coefficient k of cognitive function failure risk by causal relationship analysis based on the DEMATEL decision laboratory method.

[0033] a. Organize expert evaluation to form a direct impact matrix D Expert evaluation was employed to determine the causal relationships and initial weights among information elements. Specifically, the Delphi method was used to score the pairwise influence between information elements, with a scoring scale of 0, 1, 2, and 3, corresponding to "no influence," "low influence," "high influence," and "very high influence," respectively. A direct influence matrix was then generated based on the scoring results. .

[0034] (3); In the formula, The value is determined based on expert scoring. n represents the total number of information elements.

[0035] b. Direct Influence Matrix After normalization, the standardized influence matrix M is obtained.

[0036] This will directly affect the matrix. Normalization is performed to ensure that the matrix is ​​directly affected. The value is within the interval [0,1]. The normalization formula is: (4); The overall relationship matrix is ​​obtained by calculating the standardized influence matrix: (5); In the formula It is the identity matrix; Overall relational matrix.

[0037] In the Dematel framework, the status and importance of an element are typically characterized by two metrics: "centrality" and "causation." Centrality (Z) is defined as the sum of an element's "influence" and "affectedness," used to measure its overall role and position in the system. Causation (V) is defined as "influence" minus "affectedness," used to determine whether an element is more of a "cause" or a "result." If >0, then this element mainly affects other elements. If the value is less than 0, the element is mainly affected by other elements. It can be calculated using equation (6).

[0038] (6); In the formula: Indicates the degree of influence of element i; This indicates the degree to which element i is affected; It represents the centrality of element i in the system. The larger the value, the more it is related to other factors. This represents the tendency (causality) of element i to influence other elements; n is the sample size. Represents the overall relation matrix Some elements in.

[0039] Expert evaluation was used to score the impact of each factor of the Scenario-Environment Performance Condition (CSEPC) on the four cognitive functions (perception, cognition, decision-making, and reaction) of the submariner. The "normal" level was set as the weighted baseline (1), and the other levels were adjusted accordingly. As shown in Table 2, the scores of the impact of each Scenario-Environment Performance Condition (CSEPC) on cognitive function were obtained. Based on the scores of centrality, causality, and situational performance conditions (CSEPC) on the degree of impact on cognitive function, the risk coefficient k of cognitive function failure is obtained.

[0040] (7); in This represents the coefficient of influence of situational performance conditions (CSEPC factor) on the risk of cognitive functioning errors. Centrality of representation; Indicates the degree of causality of element 1. The score represents the degree of influence of the CSEPC factor on cognitive function.

[0041] Table 2. Correspondence Table of 5-Level Scales for Evaluating the Impact Coefficient of Cognitive Failure Risk

[0042] S102. Using fuzzy trigonometric functions to determine cognitive function risk assessment factors. .

[0043] This invention uses fuzzy trigonometric functions to characterize uncertain fuzzy sets (see...). Figure 6 Based on the CSEPC performance condition assessment criteria, the fuzzy set value range of the language items for risk expectation is defined: the range corresponding to "reduce" is [0,50], the range corresponding to "not significant" is [10,90], and the range corresponding to "increase" is [50,100].

[0044] This invention uses 10 CSEPC factors as input. The risk level of each CSEPC factor is characterized within a fuzzy set. Given that each CSEPC factor may have three types of impact on human reliability—increasing, neutral, or decreasing—a membership function can be constructed for 9 of the CSEPC factors to determine the output fuzzy set.

[0045] ; ; (8); In the formula, To reduce risk, To be neutral on the impact of risk, To increase risk.

[0046] Equation (8) shows the membership function of CSEPC. Based on the membership function, the fuzzy output of the impact of CSEPC factors on human reliability can be obtained, that is, the weights of the six cases in which the impact of CSEPC factors on human reliability occurs, as shown in Equation (9).

[0047] (9); To simplify the fuzzy assessment process, an adjustment rule for the risk assessment factor γ can be obtained based on six weighting scenarios, ultimately yielding the cognitive function risk assessment factor. The adjustment rules are as follows: (10); In the formula As a risk assessment factor for cognitive function, The average score given by the experts.

[0048] S103. Based on the risk factor weights of scenario performance conditions The comprehensive situational risk index S is obtained from three indicators: the impact coefficient of cognitive function error risk k, and the cognitive function risk judgment factor γ. (11); In the formula, It is the first A comprehensive index of situational risk for cognitive functions. It is the first Item scenario performance condition risk factor weights, It is the first Scenario, environment, and performance conditions for the first The impact coefficient of the risk of cognitive function failure. It is the first Cognitive function risk assessment factors for situational environmental performance conditions. Reduce cognitive function risks No risk of affecting cognitive function Increased cognitive function risk.

[0049] S104. Determine the risk control model and error probability range based on the scenario risk comprehensive index S.

[0050] Based on the scenario risk comprehensive index S, a correspondence model between the scenario environmental performance conditions and risk control modes of manned submersibles is established, and a coordinate system [X,Y]=[∑ 提升 , ∑ 降低 Let X represent the cumulative amount of the enhancing effect and Y represent the cumulative amount of the reducing effect, expressed as: (12); In the formula ∑ 降低 This represents the number of CSEPCs that can reduce the risk of personnel operational performance, ∑ 提升 The number of CSEPCs that indicate an increase (exacerbation) in personnel operational performance risk.

[0051] like Figure 5 As shown, this method can refer to the scenario risk composite index to determine the risk control model: S∈[−8,−4] is strategic; S∈[−3,1] is tactical; S∈[2,4] is opportunistic; when S∈[6,10], if the number of factors reducing performance risk ∑ 降低 ≥1 represents an opportunity-based risk factor; if the number of performance risk factors ∑ is reduced... 降低 =0 indicates a chaotic situation. Specific scenario risk indices and cognitive risk control models can be found in Table 3.

[0052] Table 3 Comparison between Situational Risk Comprehensive Index and Cognitive Risk Control Model

[0053] S2. Prediction of the probability of cognitive errors of submariners based on the improved CREAM extension method (1) First, based on the situational environment performance condition (CSEPC), an improved CREAM risk prediction model is constructed to obtain the probability of cognitive error of the submariner. Adaptive fuzzy logic is introduced to dynamically correct the probability of cognitive error according to the complexity of the task and changes in the environment. Specifically:

[0054] Using the situational risk composite index S to describe the change in cognitive error probability (CFP), a relationship between cognitive error probability (CFP) and its basic error probability can be established. The quantitative relationship between the scenario risk composite index S (improved CREAM risk prediction model), and the probability of cognitive error: (13); set up The constant is used. Based on the existing static scenario risk comprehensive index S, an adaptive fuzzy logic method is introduced to consider the dynamic impact of task complexity and environmental changes. Fuzzy membership functions for task complexity C and environmental change E are defined. and The assessment of the probability of cognitive error is adjusted in real time. The dynamically adjusted probability of cognitive error (CFP*) can be expressed as: ; (14); in, It is an adaptive fuzzy correction factor used to dynamically correct static error probabilities. and These represent the fuzzy membership degree of the task complexity and environmental changes, respectively, calculated based on real-time data. To adjust the coefficient, is the sensitivity coefficient of task complexity to error probability, representing the weight of the amplification effect of complex tasks on the error probability. This is the sensitivity coefficient of environmental change to the probability of error, representing the weight of the amplification effect of environmental fluctuations on the probability of error.

[0055] (2) To improve the basic value of the error probability To ensure structural accuracy, weights are introduced for each error mode (cognitive function), and the final basic error probability value is obtained by comprehensively considering the weights of the probabilities of each error mode. The formula is as follows: (15); In the formula The basic error probability of cognitive behavior; Weights for perceptual and cognitive functions; For cognitive function weights; Weights for decision-making cognitive functions; Weights reflecting cognitive function; Let be the weight of the i-th error pattern under cognitive function G; Let be the basic error probability of the i-th term under cognitive function G; Let J be the weight of the j-th error pattern under cognitive function R; Let be the basic error probability of the j-th term under cognitive function R; For the weight of the i-th error pattern under cognitive function J; Let J be the basic error probability of the j-th term under cognitive function J; The weight of the h-th error pattern under cognitive function F; Let h be the basic error probability of the h-th term under cognitive function F.

[0056] Experts were invited to assess the cognitive function types and error pattern types of the target task, and the cognitive function weights and pattern weights were obtained respectively. Table 4 lists the cognitive function weights and error pattern weights.

[0057] Table 4. Weights of Cognitive Function and Error Pattern

[0058] S3. Human Factors Reliability Assessment for Dynamic Cabin Environments of Manned Submersibles Based on the cognitive error probability obtained in step S2, step S3 further introduces multi-source observation information and a dynamic correction mechanism to achieve real-time updates and environmental adaptive adjustments to the prediction results, thereby obtaining the final cognitive error probability at the task time level. First, DS evidence simplification and fusion are performed on multi-source observations of environment, behavior, and physiology to obtain robust observations, which serve as the observation input for real-time DBN updates, estimating the cognitive error probability at time t online. Then, a dynamic environmental performance correction factor is constructed to explicitly characterize the immediate impact of the environment on human performance.

[0059] S301, Multi-source information fusion based on DS evidence To reduce the uncertainty of multi-source data and improve the accuracy of evaluation, this invention uses the Dempster–Shafer evidence theory to fuse trust assignment and basic probability assignment (BPA) to output the observations at time t. ∈[0,1], Observation A higher value indicates a higher perceived risk, and is used in observation models.

[0060] (1) Single-source BPA To robustly characterize the strength of evidence and uncertainty of each information source before fusion, single-source observations are transformed into an identification framework. Basic probability assignment (BPA) for (normal | risk). Let the identification framework be applied to the first... Standardized score of each information source at time t (The larger the value, the more inclined it is to R), given the trust / reliability weight of the source. ∈[0,1], its BPA is defined as: (16); In the formula For the first The quality allocation of a proposition by an information source at time t satisfies the following condition: ; To identify the frame, N represents normal and R represents risk.

[0061] (2) Two-source DS fusion After completing the single-source BPA and trust assignment for each information source, it is necessary to fuse evidence from different sources under the same identification framework for consistency. To this end, the Dempster combination rule is adopted: first, a normalized synthesis of the two-source case is given to explicitly handle inter-source conflicts and maintain quality uniformity; then, the multi-source cases are iteratively combined sequentially to obtain a unified fusion result for subsequent probabilistic analysis and DBN observation. The Dempster combination of p fusion is: (17); In the case of multiple information sources, the data is iteratively fused sequentially to obtain the total BPA after fusion: (18); In the formula Let A, B, and C be the basic probability assignments of information sources o and p at time t; and let C be the recognition frames. A subset of represents the set of possible propositions / hypotheses; To assess the quality of proposition A after the fusion of the two information sources; Let O be the basic probability assignment of the o-th information source to the proposition set B at time t; Let p be the basic probability assignment of the p-th information source to the proposition set C at time t; The degree of conflict between the evidence from the two sources; For all S n The total BPA after fusion of information sources at time t.

[0062] (3) Probabilistic output All S are obtained from steps (1) and (2). n After the total BPA of the information sources is fused at time t, the evidence quality is transformed into probabilistic observations that can be directly utilized by the DBN. While maintaining a neutral allocation of "unknown / informationless" quality, the fusion result is mapped to observations. , The specific formula is as follows: (19); In the formula Assigning basic probabilities after fusion; For the quality of trust in risk propositions; The betting probability is obtained by distributing uncertainty equally between risk and normal. It is the probability of betting on the "risk" of the proposition; As input for subsequent DBN observations.

[0063] S302, Real-time updates based on DBN To reduce the static dependence on general performance conditions, this invention utilizes the dynamic cognitive error prior obtained from static evaluation. Building upon this foundation, a Dynamic Bayesian Network (DBN) is introduced to update the error probability in real time as the task progresses and the environment changes. The network comprises three time-rolling layers: latent variables... (Cognitive control model); Observations ; Errors or incidents. Among them... Control mode quantities corresponding to CREAM or COCOM; For observation; Indicates whether a cognitive error occurred at time t; This represents the mode conditional error rate. A first-order Markov structure is used here. → → ,and → . → The cognitive control pattern shifts over time in the temporal dimension; pattern It affects both directions simultaneously, inwards. → Whether the impact is incorrect, outward → This affects the observed data. Specifically:

[0064] (1) Transition probability (modulated by scenario / fuzzy factor) set up Let be the comprehensive risk index for the scenario at time t. , ∈[0,1] represent the fuzzy membership degrees of task complexity and environmental change, respectively. Then: (20); In the formula Let be the probability that the cognitive control mode transitions from the previous state u to the current state v at time t; u is the mode at the previous time; v is the candidate mode at the current time; w is all possible candidate modes in the denominator, used for normalization. The transition probability is time-varying. For the intercept term, These are the scenario risk composite indexes. Task complexity and fuzzy membership Fuzzy membership degree due to environmental changes The coefficients; the denominator is a normalization term, which guarantees that the sum of probabilities for all v is 1.

[0065] (2) Conditional Gaussian observation model To characterize the statistical relationship between multi-source observation and cognitive control patterns, let a given... When each observation dimension follows a Gaussian distribution and is conditionally independent, a conditional Gaussian observation model is obtained: (twenty one); In the formula Indicating in cognitive control mode Under known conditions, the observation The probability distribution; Let be the observation at time t; Control modes Next The mean and variance parameters of the dimensional features; assuming given Each dimension is independent under the given conditions.

[0066] (3) Forward recursion Given the time-varying transition probabilities and the conditional Gaussian observation model, a forward filtering recursive approach is used to update the posterior distribution of the cognitive control pattern online during task execution. The recursive relationship is as follows: (twenty two); In the formula This is a cognitive control model; and It indicates a transition from one mode in the previous moment to another mode in the current moment; To predict priors; It is a forward-posterior distribution; The T-1 submariner is in control mode at this time. The posterior probability of is derived from the forward recursion result of the previous time step.

[0067] (4) Error probability of DBN output After completing the forward recursion and obtaining the posterior distribution of each control mode Subsequently, to map the discrete cognitive control state to the overall error risk at time t, the conditional error rate of each mode is used. Perform a full probability weighting. The weighted result is the probability of cognitive error in DBN at time t, expressed as follows: (twenty three); S303, Dynamic Environmental Performance Correction Factor Construction This invention constructs a dynamic environmental performance correction factor based on a multi-parameter environmental suitability analysis method. The derivation model is described. This model takes several key dynamic cabin environmental parameters as input, including cabin temperature, humidity, oxygen concentration, carbon dioxide concentration, noise level, and lighting conditions. Through membership mapping and weighted fusion, a normalization function is constructed to correct the cognitive error probability output by the DBN.

[0068] Considering the characteristics of the submersible cabin environment and its potential impact on human factors reliability, this invention selects the following dynamic cabin environment parameters, as shown in Table 5, that significantly affect the cognitive performance of submersible pilots as evaluation criteria. Input variables.

[0069] Table 5 Dynamic Cabin Environmental Parameters

[0070] To quantitatively describe the impact of each parameter on human factors reliability, this invention uses a trapezoidal fuzzy membership function to map the suitability of each dynamic cabin environment parameter to the [0,1] interval.

[0071] For the Parameters Its membership function is defined as follows: (twenty four); In the formula: For the first Measured values ​​of dynamic cabin environmental parameters For a specific moment; For parameters The corresponding uniform membership value represents the degree of benefit to cognitive performance, ranging from [0,1]. , , , Based on the spectrum of environmental risk factors, the general medical requirements for temperature environment in working compartments (GJB 898-90), and the permissible concentrations of air components in conventional submarines (GJB11B-2012), four key thresholds are set as shown in Table 5.

[0072] Setting the first The weight of the item parameter is This weight is obtained through expert scoring and reflects the relative importance of each dynamic parameter to the overall cognitive performance of the submariner, and satisfies the following: (25); Obtain the suitability index of dynamic cabin environmental parameters. , represented as: (26); In the formula, N represents the total number of selected dynamic cabin environmental parameters; Let p be the weight of the p-th dynamic environmental parameter on the overall cognitive performance of the submariner. The cabin environment suitability index indicates the overall suitability of the current dynamic cabin environment for the overall cognitive performance of the submariner. The higher the value, the more favorable the environment.

[0073] The cabin environment suitability index Mapped to cognitive error correction factor This reflects the dynamic effect of environmental changes on the probability of errors. Because A higher value indicates a more favorable environment for cognitive performance and a lower probability of cognitive errors. Should with They exhibit an inverse relationship. This invention employs the following linear mapping function to reflect the amplifying effect of environmental conditions on the probability of cognitive errors: (27); constant It was decided that under the most unfavorable environmental conditions (i.e.) =0), the maximum amplification factor of the probability of cognitive error. According to the comparison table of situational risk comprehensive index and cognitive risk control model, the work efficiency value is [0.00005, 1.0], and the difference between the minimum and maximum value is 20,000 times. Therefore, this invention sets the probability of cognitive error to be amplified by 20,000 times under the most unfavorable environmental conditions. =log10(20000)=4.3, therefore we get: (28); S304, the Contextual Environment Risk Performance Condition (CSEPC) can clearly describe the relationship between the contextual environment and cognitive behavior during submariner operations. However, since this method is a static condition for personnel performance in a contextual environment, it ignores the impact of changing contextual environments on personnel's operational capabilities. Therefore, based on equation (14) in step S2, this invention uses DBN for real-time updates and adjusts the cognitive error correction factor accordingly. The final probability of cognitive error is obtained as follows: (29); In the formula, This represents the final cognitive error probability at time t (used for subsequent evaluation of the series-parallel system and task-level assessment). This represents the probability of cognitive error at time t obtained by forward recursion of DBN; This is a dynamic environmental performance correction factor.

[0074] For dynamic cabin environments, firstly, simplified fusion of DS evidence (trust assignment and BPA) is used to generate robust one-dimensional observations even with noisy, missing, or conflicting data, reducing the interference of uncertainty on judgment. Secondly, real-time updates based on DBN are adopted to transform multi-source information during the mission into dynamic estimates of the submariner's cognitive state and error risk, improving the response speed to changes in the situation. Thirdly, through dynamic environmental performance correction factors, the favorable / adverse effects of the cabin environment are directly mapped to the risk assessment results, reflecting the immediate impact of the environment on human factors performance. These three aspects work synergistically to ensure that the assessment results retain the structured advantages of the CREAM system while possessing online update capabilities and engineering feasibility; simultaneously, it significantly reduces reliance on static values ​​of the general CPC, improving the accuracy, robustness, and timeliness of human factors reliability assessment in complex deep-sea missions.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic and static collaborative evaluation method for the human factors reliability of submariners based on an improved CREAM framework, characterized in that, Includes the following steps: S1. Based on the constructed scenario-environment risk factor spectrum, determine the scenario-environment performance condition CSEPC, use the decision laboratory method to obtain the cognitive function error risk impact coefficient k, and determine the cognitive function risk judgment factor. Then, the scenario risk comprehensive index S is calculated, and the risk control model and error probability range are obtained based on the scenario risk comprehensive index S. S2. Based on the scenario risk comprehensive index S, construct an improved CREAM risk prediction model to obtain the probability of cognitive error, and introduce adaptive fuzzy logic to correct the error probability. S3. Based on the cognitive error probability obtained in step S2, multi-source observation information is introduced for fusion, and a dynamic Bayesian network (DBN) is used for real-time updates, combined with a dynamic environmental performance correction factor. This allows us to obtain the final cognitive error probability at the task moment level.

2. The dynamic and static collaborative assessment method for the human factor reliability of submariners based on the improved CREAM as described in claim 1, characterized in that, Step S1 specifically includes: S101. Construct a spectrum of scenario-environment risk factors and use information entropy to calculate the weights of scenario performance condition risk factors in the spectrum of scenario-environment risk factors. The situational environment performance condition (CSEPC) was screened and determined by assessing the impact of four cognitive functions on risk factors in the risk factor spectrum of the situational environment, and the risk impact coefficient k of cognitive function failure was obtained by using the decision laboratory method. S102. Using fuzzy trigonometric functions to determine cognitive function risk assessment factors. ; S103. Based on the risk factor weights of scenario performance conditions The risk impact coefficient k of cognitive function failure and the comprehensive risk index S of the cognitive function risk assessment factor γ are obtained from the scenario risk: ; In the formula, It is the first A comprehensive index of situational risk for cognitive functions. It is the first Item scenario performance condition risk factor weights, It is the first Scenario, environment, and performance conditions for the first The impact coefficient of the risk of cognitive function failure. It is the first Cognitive function risk assessment factors for performance conditions in specific scenarios and environments; S104. Based on the scenario risk comprehensive index S, establish a correspondence model between the scenario environment performance conditions of manned submersibles and risk control modes, and determine the risk control mode and the error probability range.

3. The dynamic and static collaborative evaluation method for the human factor reliability of submariners based on the improved CREAM as described in claim 2, characterized in that, Step 101 specifically includes: S1011. Based on the characteristics of complex scenarios and environments for manned submersibles, the risk factors of the scenario and environment are processed and divided into five categories: organizational factors, task factors, personal factors, human-machine factors and environmental factors, and a spectrum of risk factors for complex scenarios and environments of manned submersibles is constructed. S1012. Using information entropy, calculate the weights ω of each situational performance condition risk factor for each situational performance condition risk factor and the information elements of the four core cognitive functions. ; In the formula This represents the total number of information elements in the system. For information elements Normalized weights; For information elements The entropy value, ; For information elements The characteristics; For each information element The probability of the state occurring, Indicates the possible states of an information element; Based on the weight ω of each scenario performance condition risk factor, the secondary indicators of the scenario performance condition risk factors are ranked to obtain the scenario environment performance condition CSEPC. S1013. The Delphi method is used to score the degree of influence between each pair of information elements, and the direct influence matrix is ​​obtained based on the scoring results. The normalized influence matrix M is obtained by performing normalization. ; The overall relationship matrix is ​​obtained by calculating the standardized influence matrix: ; In the formula It is the identity matrix; Overall relational matrix; In the decision laboratory method, centrality is used as the basis for decision-making. and causal degree As two indicators, the calculation formula is: ; In the formula: Indicates the degree of influence of element i; This indicates the degree to which element i is affected; It represents the centrality of element i in the system. The larger the value, the more it is related to other factors. This represents the tendency of element i to influence other elements; n is the sample size. Represents the overall relation matrix Some elements in; Expert evaluation was used to score the impact of the Scenario-Environment Performance Conditions (CSEPC) on the four cognitive functions of submariners, resulting in a score for the impact of each CSEPC condition on cognitive function. This allows for the determination of the impact coefficient of cognitive impairment risk. ; 。 4. The dynamic and static collaborative assessment method for the human factor reliability of submariners based on the improved CREAM as described in claim 2, characterized in that, Step 102 specifically involves: The risk level of each scenario-environment performance condition (CSEPC) is characterized within a fuzzy set. Each scenario-environment performance condition (CSEPC) has three types of impact on human reliability: increasing, neutral, and decreasing. Corresponding membership functions are then constructed as follows: ； ； ； In the formula, To reduce risk, To be neutral in terms of risk impact, To increase risk; Based on membership functions, obtain the weights of the impact of situational performance conditions (CSEPC) on human reliability: ; The cognitive function risk assessment factor is ultimately obtained based on the weights. : ; In the formula As a risk assessment factor for cognitive function, The average score given by the experts.

5. The dynamic and static collaborative assessment method for the human factor reliability of submariners based on the improved CREAM as described in claim 1, characterized in that, Step S2 is as follows: (1) Introduce adaptive fuzzy logic method, task complexity C and environmental change E fuzzy membership function into the scenario risk comprehensive index S. and Obtain the probability of cognitive error (CFP*); ； ； in, It is an adaptive fuzzy correction factor; To adjust the coefficient, This is the sensitivity coefficient of task complexity to error probability; This is the sensitivity coefficient of the probability of error to environmental changes; (2) Based on the weight values ​​of each cognitive function, and considering the weights of the probability of occurrence of each error pattern, obtain the final basic value of the error probability: ; In the formula The basic error probability of cognitive behavior; Weights for perceptual and cognitive functions; For cognitive function weights; Weights for decision-making cognitive functions; Weights reflecting cognitive function; Let be the weight of the i-th error pattern under cognitive function G; Let be the basic error probability of the i-th term under cognitive function G; Let J be the weight of the j-th error pattern under cognitive function R; Let be the basic error probability of the j-th term under cognitive function R; For the weight of the i-th error pattern under cognitive function J; Let J be the basic error probability of the j-th term under cognitive function J; The weight of the h-th error pattern under cognitive function F; Let h be the basic error probability of the h-th term under cognitive function F.

6. The dynamic and static collaborative evaluation method for the human factor reliability of submariners based on the improved CREAM as described in claim 1, characterized in that, Step S3 is as follows: S301. Based on the DS evidence theory, trust assignment and basic probability assignment are fused to obtain the observations at time t. ; S302. Construct a dynamic Bayesian network DBN and update the probability of cognitive error in real time based on the dynamic Bayesian network DBN. S303. Based on the multi-parameter environmental suitability analysis method, a dynamic environmental performance correction factor was constructed. The derivation model is as follows: a normalization function is constructed through membership mapping and weighted fusion to correct the cognitive error probability output by DBN. S304. The final cognitive error probability is calculated by combining the dynamic environment performance correction factor and the cognitive error probability output by DBN.

7. The dynamic and static collaborative evaluation method for the human factor reliability of submariners based on the improved CREAM as described in claim 6, characterized in that, Step S301 specifically includes: (1) Convert single-source observations into identification frameworks Basic probability assignment (BPA); Suppose the identification frame is for the first... Standardized score of each information source at time t Given the trust or reliability weight of the information source ∈[0,1], its BPA is defined as: ; In the formula For the quality allocation of propositions by information source s at time t, satisfying ; To identify the framework, N represents normal and R represents risk; (2) The Dempster combination rule is used to analyze the two information sources. p-fusion: ; In the case of multiple information sources, the data is iteratively fused sequentially to obtain the total BPA after fusion: ; In the formula Let A, B, and C be the basic probability assignments of information sources o and p at time t; and let C be the recognition frames. A subset of; To determine the quality of proposition A after the fusion of the two sources; Let O be the basic probability assignment of the o-th information source to the proposition set B at time t; Let p be the basic probability assignment of the p-th information source to the proposition set C at time t; The degree of conflict between the evidence from the two sources; For all The total BPA after fusion of information sources at time t; (3) Map the fusion results to observations , The specific formula is as follows: ; In the formula For the quality of trust in risk propositions; The betting probability is obtained by distributing uncertainty equally between risk and normal. It is the probability of betting on the proposition "risk".

8. The dynamic and static collaborative assessment method for the human factor reliability of submariners based on the improved CREAM as described in claim 6, characterized in that, Step S302 specifically includes: (1) The dynamic Bayesian network contains three time-rolling layers: latent variables ; Observation ; Error incident; (2) Let Let be the comprehensive risk index for the scenario at time t. , ∈[0,1] represent the fuzzy membership degrees of task complexity and environmental change, respectively. Then: ; In the formula Let v be the probability that the cognitive control mode transitions from the previous state u to the current state v at time t. u represents the pattern from the previous time step; v represents the candidate patterns from the current time step; w represents all candidate patterns in the denominator, used for normalization. The time-varying transition probability; For the intercept term, These are the scenario risk composite indexes. Task complexity and fuzzy membership Fuzzy membership degree due to environmental changes The coefficient; (3) Given When each observation dimension follows a Gaussian distribution and is conditionally independent, a conditional Gaussian observation model is obtained: ; In the formula Indicating in cognitive control mode Under known conditions, the observation The probability distribution; Control modes Next The mean and variance parameters of the dimensional features; (4) Given the time-varying transition probability and the conditional Gaussian observation model, the posterior distribution of the cognitive control mode is updated recursively using forward filtering; the recursive relationship is as follows: ; In the formula This is a cognitive control model; and It indicates a transition from one mode in the previous moment to another mode in the current moment; To predict priors; It is a forward-posterior distribution; The T-1 submariner is in control mode at that moment. The posterior probability of ; (5) Conditional error rate of each mode Perform a full probability weighting; the weighted result is the probability of cognitive error of DBN at time t: .

9. The dynamic and static collaborative evaluation method for the human factor reliability of submariners based on the improved CREAM as described in claim 6, characterized in that, Step S303 specifically includes: Trapezoidal fuzzy membership functions are used to map the suitability of each dynamic cabin environment parameter to the [0,1] interval; for the th Parameters Its membership function is: ; In the formula: For the first Measured values ​​of dynamic cabin environmental parameters For a specific moment; For parameters The corresponding uniformized membership value ranges from [0,1]. , , , Based on the spectrum of environmental risk factors, as well as the general medical requirements for temperature environment in working compartments (GJB 898-90) and the permissible concentration of air components in conventional submarines (GJB11B-2012), these four key thresholds are set. Setting the first The weight of the item parameter is This weight is obtained through expert scoring and reflects the relative importance of each dynamic parameter to the overall cognitive performance of the submariner, and satisfies the following: ; Obtain the suitability index of dynamic cabin environmental parameters. , is represented as: ; In the formula, N represents the total number of selected dynamic cabin environmental parameters; Let p be the weight of the p-th dynamic cabin environment parameter on the overall cognitive performance of the submariner; The suitability index of dynamic cabin environmental parameters Mapped to cognitive error correction factor : ; constant This determines the maximum amplification factor of the probability of cognitive error under the most unfavorable environmental conditions.

10. The dynamic and static collaborative assessment method for the human factor reliability of submariners based on the improved CREAM according to claim 6, characterized in that, Step S304 specifically includes: Based on cognitive error correction factor The final probability of cognitive error is obtained as follows: ; In the formula, This represents the probability of a final cognitive error at time t; This represents the probability of cognitive error at time t obtained by forward recursion of DBN; This is a dynamic environmental performance correction factor.

Images