Multi-level human factor association quantitative mining method for unsafe events

Abstract

The invention relates to an unsafe event multi-level human factor association quantitative mining method, which comprises the following steps of: performing multi-factor coupling analysis based on an SHELL model on an event survey report; carrying out human factor analysis based on an HFACS model on an analysis result of the SHELL model; performing grey relational degree analysis based on GRA on an analysis result of the HFACS model; carrying out vulnerability analysis based on an N-K model on an analysis result of the HFACS model; and combining a grey correlation degree analysis result with a vulnerability analysis result to obtain an unsafe event multi-level human factor correlation quantitative mining result. According to the method, a civil aviation field unsafe event human factor quantitative analysis process based on the event survey report is established, the method is a feasible new method for mining and quantifying deep reasons causing unsafe behaviors, and objective quantitative theoretical reference is provided for formulating targeted risk avoidance measures.

Description

Technical Field

[0001] The present invention relates to the field of failure mode analysis and cause mining of unsafe events in civil aviation, and in particular to a method for quantitative mining of multi-level human factor associations in unsafe events. Background Art

[0002] Civil aviation safety is one of the most important concerns in the civil aviation field, and is particularly valued by relevant units in the civil aviation industry. When an accident or unsafe incident occurs, both airlines and relevant organizations will conduct accident and incident investigations, issue accident and incident investigation reports, summarize the causes of the incident, and optimize safety management.

[0003] Among the many risks to civil aviation safety and the various causes of civil aviation unsafe incidents, human factors are one of the key factors and the intersection of research in various directions in the field of civil aviation. Factor research focuses on the analysis, evaluation and impact assessment of human factors. There is less research on the relationship between unsafe behaviors and hidden failures, a lack of quantitative research on the correlation between levels, and a low utilization rate of incident investigation reports.

[0004] Refine the correlation between unsafe behaviors and coupled influencing factors at all levels, in order to provide objective and quantitative theoretical reference for the formulation of human factor prevention and control measures in civil aviation safety management. This will target civil aviation practitioners. Combining theory with practice, comprehensively enhance the crew's safety awareness and emergency response capabilities, further improve the civil aviation safety system for civil aviation production and operation units, implement the responsibility system, and prevent major risks. Summary of the invention

[0005] The purpose of the present invention is to propose a method for quantitative mining of multi-level human factor associations in unsafe incidents to solve the problems existing in the above-mentioned prior art, and to establish a quantitative analysis process of human factors in unsafe incidents in the civil aviation field based on event investigation reports. It is a practical new method for mining and quantifying the deep-seated causes of unsafe behaviors, and provides an objective and quantitative theoretical reference for formulating targeted risk avoidance measures.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for quantitative mining of multi-level human factor associations in unsafe events, including:

[0008] Conduct multi-factor coupling analysis based on the SHELL model for the incident investigation report;

[0009] Conduct human factor analysis based on HFACS model for the analysis results of SHELL model;

[0010] The analysis results of the HFACS model are analyzed by grey relational analysis based on GRA;

[0011] Conduct vulnerability analysis based on the NK model for the analysis results of the HFACS model;

[0012] Combine the grey correlation analysis results with the vulnerability analysis results to obtain the multi-level human factor correlation quantitative mining results of unsafe events. Combine the grey correlation analysis results with the security system vulnerability analysis results to obtain the multi-level human factor correlation quantitative mining results of unsafe events.

[0013] Optionally, a multi-factor coupling analysis based on the SHELL model includes:

[0014] The investigation report of the incident analyzes the causes of serious transport aviation incidents based on single factors of preset angles; wherein the preset angles include: software, hardware, environment, people, and other people; the people refer to the crew members, the other people refer to other people other than the pilots, and the environment includes: meteorological environment, airport terrain environment, airspace traffic environment, cockpit environment, company operating environment, and humanistic and social environment;

[0015] For the analysis results of single factors, multiple factors are superimposed to obtain multi-factor events;

[0016] For the multi-factor event, a failure mode analysis is performed under the coupling of multiple factors.

[0017] Optionally, performing a human factors analysis based on the HFACS model includes:

[0018] Based on the analysis results of the SHELL model, the HFACS model is used to analyze the influencing factors at four levels: unsafe behaviors of human factors, premises of unsafe behaviors, unsafe supervision and organizational influence. The grey correlation analysis method is used to explore the influence of each underlying factor on the occurrence of unsafe behaviors; the unsafe behavior index is constructed; the top three main influencing factors leading to habitual violations and unsafe behaviors and the main influencing factors leading to accidental violations are obtained.

[0019] Optionally, constructing the unsafe behavior indicator includes:

[0020] The unsafe behavior level is taken as the dependent variable; the errors and violations classified by unsafe behavior are taken as dependent variables respectively; skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations are taken as dependent variables respectively; the 10 influencing factors of resource management, organizational atmosphere, organizational process, insufficient supervision, inappropriate operation plan, failure to correct existing problems, supervision violations, environmental factors, operator status, personal preparation and CRM refined from the three levels of unsafe behavior premise, unsafe supervision and organizational influence are taken as independent variables; based on the dependent variable and the independent variable, the unsafe behavior index is constructed.

[0021] Optionally, performing a grey relational analysis based on GRA includes:

[0022] Identify, refine and categorize the specific causes of unsafe incidents based on the pilot's direct causes, indirect causes, root causes and root causes, and according to the detailed descriptions in the unsafe incident report;

[0023] Based on the unsafe behavior indicators, the secondary influencing factors and their detailed reasons at four levels, namely, unsafe behavior, unsafe behavior premise, unsafe supervision and organizational influence, are counted using the 1 / 0 system;

[0024] Calculate the grey correlation degree for the statistical results;

[0025] According to the results of grey correlation calculation, the first-level, second-level and third-level indicators are analyzed respectively to analyze the deep-seated important factors that lead to the occurrence of unsafe behaviors; the first-level indicator is unsafe behavior, the second-level indicator is errors and violations, and the third-level indicators are skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations;

[0026] On the basis of grey correlation analysis, association rules are used to calculate and screen the support, confidence and improvement of ten influencing factors at three levels, namely, the premise of unsafe behavior, unsafe supervision and organizational influence, which lead to the occurrence of unsafe behavior; among them, the ten influencing factors include: environmental factors, operator status, personnel factors, insufficient supervision, inappropriate operation plan, uncorrected problems, supervision violations, resource management, organizational atmosphere and organizational process.

[0027] Optionally, calculating the grey relational degree includes:

[0028] Determine the reference sequence and comparison sequence; the independent variables are the three levels of unsafe behavior premise, unsafe supervision and organizational influence and their underlying influencing factors, that is; the dependent variable is the unsafe behavior and its underlying influencing factors, that is, y′ j (r);

[0029] Standardization processing: taking time as the third dimension to calculate the grey relational analysis:

[0030]

[0031] Where n represents the number of time periods divided according to the event distribution; r represents the time period; i and j represent the factors in the independent variable and the dependent variable respectively; xi(r) and yj(r) represent the original values ​​of the independent variable and the dependent variable respectively, and x'i(r) and y'j(r) represent the standardized values ​​of the independent variable and the dependent variable respectively;

[0032] Compute the absolute difference of the sequence:

[0033] Δ ij(r)＝ x′ i (r)-y′ j (r)

[0034] Δ max (r) = maxΔ ij (r)

[0035] Δ min (r) = minΔ ij (r)

[0036] Among them, Δ ij (r) represents the difference between the independent variable and the dependent variable; Δ max (r) represents the maximum value among the differences; Δ min (r) represents the minimum value among the differences;

[0037] Calculate the correlation coefficient:

[0038]

[0039] Among them, λij(r) represents the correlation coefficient; ρ represents the resolution coefficient, which is generally taken as 0.5;

[0040] (5) Calculate the degree of association;

[0041]

[0042] Among them, γij represents the correlation degree.

[0043] The analysis results of the SHELL model are used to conduct a security system vulnerability analysis based on the NK model for these eleven elements: factors, software, hardware, cockpit environment, airport terrain environment, company operating environment, human and social environment, meteorological environment, airspace traffic environment, crew, and other people.

[0044] The vulnerability analysis of security systems based on the NK model includes:

[0045] Determine the number of risk elements N and the number of interdependent elements K;

[0046] The analysis results of the HFACS model are used to conduct single, double and multi-factor vulnerability coupling analysis, and the NK model is used to conduct single, double and multi-factor vulnerability coupling analysis of risk factors in the subway construction safety system;

[0047] Calculate the risk coupling probability and risk value, and evaluate the impact of different coupling factors on the total risk based on the risk occurrence probability and risk value under different coupling conditions;

[0048] Combined with the vulnerability theory, the degree of factor coupling is used to measure the size of the system vulnerability.

[0049] Optionally, the method further includes: performing a Bayesian network analysis based on BN on the dependency relationship result of the NK model; including:

[0050] Identify risk events that need to be predicted and evaluated, and identify all factors that affect risk events;

[0051] According to the dependencies between risk factors, the structure of Bayesian network is constructed;

[0052] Use Bayes' theorem to calculate the posterior probability of risk events given evidence;

[0053] Interpret the results of the Bayesian network analysis, including the relevance and importance of risk factors.

[0054] The beneficial effects of the present invention are:

[0055] The present invention first conducts a multi-factor coupling analysis based on the SHELL model on the incident investigation report; then conducts a human factor analysis based on the HFACS model on the analysis results of the SHELL model; secondly, conducts a gray correlation analysis based on GRA on the analysis results of the HFACS model; then conducts a safety system vulnerability analysis based on the NK model on the analysis results of the SHELL model; finally, the gray correlation analysis results and the safety system vulnerability analysis results are combined to obtain the multi-level human factor correlation quantitative mining results of unsafe events. The present invention establishes a quantitative analysis process of human factors in unsafe events in the civil aviation field based on incident investigation reports, which is a practical new method for mining and quantifying the deep-seated reasons that lead to the occurrence of unsafe behaviors, and provides an objective and quantitative theoretical reference for formulating targeted risk avoidance measures. The present invention refines the correlation between unsafe behaviors and coupling influencing factors at all levels, and provides an objective and quantitative theoretical reference for formulating human factor prevention and control measures in civil aviation safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0057] Figure 1 This is a flow chart of a multi-level human factor coupling correlation quantification model based on HFACS-GRA according to an embodiment of the present invention;

[0058] Figure 2 A distribution diagram of severe symptom types according to an embodiment of the present invention;

[0059] Figure 3 A statistical diagram of the number of factors affecting severe symptoms according to an embodiment of the present invention;

[0060] Figure 4 A relationship diagram of failure modes under the coupling of multiple factors according to an embodiment of the present invention;

[0061] Figure 5 This is a distribution diagram of the correlation degree of unsafe behaviors according to an embodiment of the present invention;

[0062] Figure 6 This is a distribution diagram of error correlation under the coupling of multiple factors in an embodiment of the present invention;

[0063] Figure 7 A distribution diagram of the degree of association of transport aviation violations according to an embodiment of the present invention;

[0064] Figure 8 A statistical chart of the total flight time of pilots according to an embodiment of the present invention;

[0065] Fig. 9 A word cloud diagram of the causes / important reasons of the core competencies of the embodiments of the present invention;

[0066] Fig.10 This is a word cloud diagram of the causes / important reasons of work style competency in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] 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.

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0070] like Figure 1 As shown, this embodiment proposes a multi-level human factor correlation quantitative mining method for unsafe events, including:

[0071] The investigation report of the incident, the report results are as follows: Figure 2 As shown, a multi-factor coupling analysis based on the SHELL model was conducted;

[0072] Conduct human factor analysis based on HFACS model for the analysis results of SHELL model;

[0073] Conduct human factor analysis based on HFACS model for the analysis results of SHELL model;

[0074] The analysis results of the HFACS model are analyzed by grey relational analysis based on GRA;

[0075] Conduct vulnerability analysis based on the NK model for the analysis results of the HFACS model;

[0076] The results of grey correlation analysis and vulnerability analysis are combined to obtain the quantitative mining results of multi-level human factor correlations in unsafe events.

[0077] Furthermore, the multi-factor coupling analysis based on the SHELL model includes:

[0078] The incident investigation report is subjected to single-factor statistical analysis based on preset angles; wherein the preset angles include: software, hardware, environment, people, and other people;

[0079] For the results of single-factor statistical analysis, multiple factors are superimposed to obtain multi-factor events;

[0080] For the multi-factor event, a failure mode analysis is performed under the coupling of multiple factors.

[0081] Specifically, in this embodiment, the cause mining analysis of unsafe incidents is carried out, and the multi-factor coupling analysis based on the SHELL model is carried out, including single-factor statistical analysis including five angles: software, hardware, environment, people, and other people.

[0082] The cause is not single, so multiple factors must be superimposed, and unsafe events are further divided into single-factor events, double-factor events, three-factor events, four-factor events, five-factor events, and six-factor events. Failure mode analysis under the coupling of multiple factors is carried out according to the events, such as Figure 3 shown.

[0083] Furthermore, the human factors analysis based on the HFACS model includes:

[0084] The HFACS model is used to refine the human factors in the unsafe events in the analysis results of the SHELL model, and statistically analyze the pilot's unsafe behavior, the premise of unsafe behavior, unsafe supervision and organizational factors from the direct cause, indirect cause, fundamental cause and root cause of the pilot, and construct the unsafe behavior index, and refine the superposition results such as Figure 4 shown.

[0085] The construction of the unsafe behavior indicator includes:

[0086] The unsafe behavior level is taken as the dependent variable; the errors and violations classified by unsafe behavior are taken as dependent variables respectively; skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations are taken as dependent variables respectively; the 10 influencing factors of resource management, organizational atmosphere, organizational process, insufficient supervision, inappropriate operation plan, failure to correct existing problems, supervision violations, environmental factors, operator status, personal preparation and CRM refined from the three levels of unsafe behavior premise, unsafe supervision and organizational influence are taken as independent variables; based on the dependent variable and the independent variable, the unsafe behavior index is constructed.

[0087] Specifically, in this embodiment, the human factors in transport aviation unsafe events are deeply analyzed according to the HFACS model, and statistical analysis is performed on the pilot's direct causes, indirect causes, fundamental causes and root causes corresponding to the pilot's unsafe behavior, the premise of the unsafe behavior, unsafe supervision and organizational factors.

[0088] In the HFACS model, the multi-level cause association mining analysis takes the unsafe behavior level as the dependent variable, the errors and violations of the unsafe behavior classification as the dependent variables, and the further refined skill errors, decision errors, perceptual errors, habitual violations and accidental violations as the dependent variables; the three levels of unsafe behavior premise, unsafe supervision and organizational influence are refined into resource management, organizational atmosphere, organizational process, insufficient supervision, inappropriate operation plan, failure to correct existing problems, supervision violations, environmental factors, operator status, personal preparation and CRM 10 influencing factors as independent variables, as shown in Table 1.

[0089] Table 1 Unsafe behavior indicators

[0090]

[0091]

[0092] Furthermore, the grey relational analysis based on GRA is conducted including;

[0093] Identify, refine and categorize the specific causes of unsafe incidents based on the pilot's direct causes, indirect causes, root causes and root causes, and according to the detailed descriptions in the unsafe incident report;

[0094] Based on the unsafe behavior indicators, the secondary influencing factors and their detailed reasons at four levels, namely, unsafe behavior, unsafe behavior premise, unsafe supervision and organizational influence, are counted using the 1 / 0 system;

[0095] Calculate the grey relational degree;

[0096] According to the results of grey correlation calculation, the first-level, second-level and third-level indicators are analyzed respectively to analyze the deep-seated important factors that lead to the occurrence of unsafe behaviors; the first-level indicator is unsafe behavior, the second-level indicator is errors and violations, and the third-level indicators are skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations;

[0097] On the basis of grey correlation analysis, association rules are used to calculate and screen the support, confidence and promotion of 10 influencing factors at three levels: premise of unsafe behavior, unsafe supervision and organizational influence, which lead to the occurrence of unsafe behavior.

[0098] Specifically, in this embodiment, based on the HFACS model, combined with GRA analysis, the shortcomings of the HFACS model, which lack quantitative calculation and inter-level correlation analysis, are compensated, and the importance of various factors causing unsafe behaviors in transport aviation unsafe incidents is further identified.

[0099] The grey correlation analysis method is used to explore the influence of each underlying factor on the occurrence of unsafe behaviors. The steps are as follows:

[0100] Step 1: Refine the cause classification. Based on the four levels of causes in the HFACS model, identify, refine and classify the specific causes of the unsafe incident according to the detailed description in the unsafe incident report.

[0101] Step 2: Detailed cause statistical analysis. According to the analysis results of the HFACS model, the secondary influencing factors and their detailed causes at four levels, namely, unsafe behavior, premise of unsafe behavior, unsafe supervision and organizational influence, are statistically analyzed using the 1 / 0 system. If a certain cause leads to an unsafe incident, it is recorded as 1, and if it does not occur, it is recorded as 0.

[0102] Step 3: Grey correlation calculation, the calculation process is as follows

[0103] (1) Determine the reference sequence and the comparison sequence. The reference sequence is a sequence that reflects the characteristics of system behavior, i.e., the dependent variable. The comparison sequence is a sequence composed of factors that affect system behavior, i.e., the independent variable. In this paper, the independent variables are the three levels of unsafe behavior premise, unsafe supervision, and organizational influence and their underlying influencing factors, i.e., y′; the dependent variable is the unsafe behavior and its underlying influencing factors, i.e., y′ j (r).

[0104] (2) Standardization. The data is normalized using the mean method. The purpose is to remove the dimension of the variable and to reduce the range of the variable to simplify the calculation. This method uses time as the third dimension to calculate the grey relational analysis.

[0105]

[0106]

[0107] Where n represents the number of time periods divided according to the event distribution; r represents the time period; i and j represent the factors in the independent variable and the dependent variable respectively; xi(r) and yj(r) represent the original values ​​of the independent variable and the dependent variable respectively, and x'i(r) and y'j(r) represent the standardized values ​​of the independent variable and the dependent variable respectively.

[0108] (3) Calculate the absolute difference of the sequence.

[0109] Δ ij(r)＝ x′ i (r)-y′ j (r)

[0110] Δ max (r) = maxΔ ij (r)

[0111] Δ min (r) = minΔ ij (r)

[0112] Where Δ ij (r) represents the difference between the independent variable and the dependent variable; Δ max (r) represents the maximum value among the differences; Δ min (r) represents the minimum value among the differences.

[0113] (4) Calculate the correlation coefficient.

[0114]

[0115] Where λij(r) represents the correlation coefficient; ρ represents the resolution coefficient, which is generally taken as 0.5.

[0116] (5) Calculate the correlation.

[0117]

[0118] In the formula, γij represents the correlation degree.

[0119] According to the calculation steps of grey correlation analysis, the correlation between the three first-level factors of unsafe behavior premise, unsafe supervision and organizational influence, resource management, organizational atmosphere, organizational process, insufficient supervision, inappropriate operation plan, failure to correct existing problems, supervision violation, environmental factors, operator status, personal preparation and crew resource management (Crew Resource Management, CRM) and the two first-level factors of unsafe behavior, error and violation, and the five second-level factors of skill error, decision error, perceptual error, habitual violation and accidental violation is calculated in turn.

[0120] Step 4: Inter-level correlation analysis. According to the results of the grey correlation calculation, analyze the first-level indicators (unsafe behavior), second-level indicators (errors, violations), and third-level indicators (skill errors, decision errors, perceptual errors, habitual violations, and accidental violations) respectively, and analyze the deep-level important influencing factors that lead to the occurrence of unsafe behaviors.

[0121] Then, on the basis of grey correlation analysis, multi-level factor strong correlation rule analysis was conducted, and correlation rules were used to calculate and screen the support, confidence and improvement of 10 influencing factors at three levels: premise of unsafe behavior, unsafe supervision and organizational influence, which lead to the occurrence of unsafe behavior.

[0122] Confidence It represents the probability of deriving Y from the association rule (X, Y) when the prerequisite X occurs, that is, the probability of unsafe behavior, error and violation when the 10 influencing factors occur.

[0123] Support It represents the probability of the item set {X, Y} appearing in the total item set, that is, the probability of the 10 influencing factors occurring simultaneously with unsafe behaviors, errors and violations in all unsafe events.

[0124] Lift It represents the ratio of the probability of containing Y under the condition of containing X to the probability of containing Y under the condition of not containing X, that is, the ratio of the probability of unsafe behavior, error and violation occurring at the same time when 10 influencing factors occur to the probability of unsafe behavior, error and violation occurring when 10 influencing factors do not occur. At the same time, a lift greater than 1 indicates a valid strong association rule, less than 1 indicates an invalid strong association rule, and equal to 1 indicates mutual independence, as shown in Table 2:

[0125] Table 2 Results of correlation analysis among factors

[0126]

[0127]

[0128] Further analysis of the correlation between unsafe behaviors, the premise of unsafe behaviors, unsafe supervision and organizational factors in the HFACS model, and the deep-seated causal factors that lead to unsafe events, errors and violations. Figure 5 , Figure 6 , Figure 7 .

[0129] Furthermore, based on the analysis results of the SHELL model, the vulnerability analysis of the security system based on the NK model includes:

[0130] Determine the number of risk elements N and the number of interdependent elements K;

[0131] The analysis results of the SHELL model are used to conduct single, double and multi-factor vulnerability coupling analysis, and the NK model is used to conduct single, double and multi-factor vulnerability coupling analysis on the risk factors in the subway construction safety system;

[0132] Calculate the risk coupling probability and risk value, and evaluate the impact of different coupling factors on the total risk based on the risk occurrence probability and risk value under different coupling conditions;

[0133] Combined with the vulnerability theory, the degree of factor coupling is used to measure the size of the system vulnerability.

[0134] Specifically, in this embodiment, the NK model is used to perform security system vulnerability analysis on the SHELL model, and the specific steps are as follows:

[0135] Step 1: Determine the number of risk elements (N) and the number of interdependent elements (K). It is necessary to identify and determine the total number of risk elements in the model and the number of interdependencies and couplings between these elements.

[0136] Step 2: Conduct single, double and multi-factor vulnerability coupling analysis, and use the NK model to conduct single, double and multi-factor vulnerability coupling analysis of risk factors in the subway construction safety system.

[0137] Step 3: Calculate the risk coupling probability and risk value, and evaluate the impact of different coupling factors on the total risk based on the risk occurrence probability and risk value under different coupling situations.

[0138] Step 4: Evaluate system vulnerability: Combined with vulnerability theory, the degree of factor coupling is used to measure the size of system vulnerability, thereby effectively increasing the objectivity and credibility of the analysis results.

[0139] Specifically, the method of this embodiment also includes: using a Bayesian network to further predict and evaluate the risk of the results of the multi-factor coupling analysis, and using prior information and posterior probability to deduce the relevance and importance of each risk factor. The specific steps are as follows:

[0140] Step 1: Determine the risk events that need to be predicted and evaluated, and identify all factors that may affect the risk events.

[0141] Step 2: Construct the structure of the Bayesian network based on the dependencies between risk factors.

[0142] Step 3: Use Bayes’ theorem to calculate the posterior probability of the risk event given the evidence.

[0143] Step 4: Interpret the results of the Bayesian network analysis, including the relevance and importance of risk factors.

[0144] In the NK model, determine the system variables (N) that you want to analyze and the constraints (K) between them. In the grey relational analysis, determine the reference sequence (parent sequence) and the comparison sequence (child sequence). Using the grey relational analysis method, calculate the correlation coefficient between each variable (or variable combination) in the NK model and the reference sequence. This can be done by calculating the absolute difference between each variable and the reference sequence at each moment, and then applying the calculation formula of the grey relational coefficient.

[0145] Conduct unsafe behavior association analysis to subdivide violation events into habitual violations and accidental violations. Conduct error association analysis to subdivide error events into skill errors, decision errors, and perceptual errors. Subdivide violation events into habitual violations and accidental violations, and find out the relationship between unsafe behaviors, violations, habitual violations, and accidental violations. Combine the results of grey correlation analysis with the simulation results of the NK model to explain the interaction and influence between variables in the system. Use the correlation degree of grey correlation analysis to verify whether the key variables and constraints in the NK model are consistent with the actual data.

[0146] This embodiment establishes a quantitative analysis process of human factors in unsafe incidents in the civil aviation field based on incident investigation reports. It is a practical new method to explore and quantify the deep-seated causes of unsafe behaviors, and provides an objective and quantitative theoretical reference for formulating targeted risk avoidance measures.

[0147] This embodiment refines the correlation between unsafe behaviors and coupling influencing factors at all levels, and provides an objective and quantitative theoretical reference for formulating human factor prevention and control measures in civil aviation safety management.

[0148] Based on the calculation results of the correlation, this embodiment analyzes the differences in the correlation between 10 deep-level influencing factors and unsafe behaviors, errors and violations, skill errors, decision errors, perceptual errors, habitual violations and accidental violations, and proposes targeted prevention and control strategies for unsafe events based on the comparison of the causes of these differences. The correlation between unsafe behaviors, the premise of unsafe behaviors, unsafe supervision and organizational factors in the HFACS model is further analyzed to explore the deep-level causal factors that lead to unsafe events, errors and violations.

[0149] Based on the HFACS model, this embodiment combines GRA analysis to make up for the lack of quantitative calculation and inter-level correlation analysis in the HFACS model, and further identifies the importance of various factors that cause unsafe behaviors in unsafe events in transport aviation. The HFACS model is used to analyze the influencing factors of the four levels of unsafe behaviors of human factors, the premise of unsafe behaviors, unsafe supervision and organizational influence in the unsafe event investigation report. The gray correlation analysis method is used to explore the influence of each underlying factor on the occurrence of unsafe behaviors. The results are as follows: Figure 8 , Fig. 9 , Fig.10 exhibit.

[0150] The NK model is used to analyze the vulnerability of the safety system of the HFACS model, and the causal mechanism of civil aviation unsafe events is analyzed: the trend of influencing factors leading to unsafe behaviors, overall errors and skill errors, and the main influencing factors are analyzed. The Bayesian network is used to construct the influence network between risk factors. In the Bayesian network, each node corresponds to a one-dimensional factor, and all sample features constitute a network structure containing multiple nodes. The relationship between these factors and project safety performance is described through the Bayesian network structure.

[0151] Through the unsafe behavior correlation analysis, the changing trends of the grey correlation between the 10 deep influencing factors and unsafe behaviors, errors, and violations are analyzed, as well as the changing trends of the grey correlation between the 10 deep influencing factors and unsafe behaviors, overall errors, and skill errors. Based on the calculation results of the correlation, the differences in the correlation between the 10 deep influencing factors and unsafe behaviors, errors and violations, skill errors, decision-making errors, perceptual errors, habitual violations, and accidental violations are analyzed. On the basis of comparing the causes of these differences, targeted prevention and control strategies for unsafe incidents are proposed.

[0152] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for quantitative mining of multi-level human factor associations in unsafe events, characterized in that: include: Conduct multi-factor coupling analysis based on the SHELL model for the incident investigation report; Conduct human factor analysis based on HFACS model for the analysis results of SHELL model; The analysis results of the HFACS model are analyzed by grey relational analysis based on GRA; Conduct vulnerability analysis based on the NK model for the analysis results of the HFACS model; The results of grey correlation analysis and vulnerability analysis are combined to obtain the quantitative mining results of multi-level human factor correlations in unsafe events.

2. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 1 is characterized in that: The multi-factor coupling analysis based on the SHELL model includes: The investigation report of the incident analyzes the causes of serious transport aviation incidents based on single factors of preset angles; wherein the preset angles include: software, hardware, environment, people, and other people; the people refer to the crew members, the other people refer to other people other than the pilots, and the environment includes: meteorological environment, airport terrain environment, airspace traffic environment, cockpit environment, company operating environment, and humanistic and social environment; For the analysis results of single factors, multiple factors are superimposed to obtain multi-factor events; For the multi-factor event, a failure mode analysis is performed under the coupling of multiple factors.

3. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 1 is characterized in that: The human factors analysis based on HFACS model includes: Based on the analysis results of the SHELL model, the HFACS model is used to analyze the influencing factors at four levels: unsafe behaviors of human factors, premises of unsafe behaviors, unsafe supervision and organizational influence. The grey correlation analysis method is used to explore the influence of each underlying factor on the occurrence of unsafe behaviors; construct unsafe behavior indicators; obtain the main influencing factors leading to habitual violations, unsafe behaviors and the main influencing factors leading to accidental violations.

4. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 3 is characterized in that: The construction of the unsafe behavior indicator includes: The unsafe behavior level is taken as the dependent variable; the errors and violations classified by unsafe behavior are taken as dependent variables respectively; skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations are taken as dependent variables respectively; the 10 influencing factors of resource management, organizational atmosphere, organizational process, insufficient supervision, inappropriate operation plan, failure to correct existing problems, supervision violations, environmental factors, operator status, personal preparation and CRM refined from the three levels of unsafe behavior premise, unsafe supervision and organizational influence are taken as independent variables; based on the dependent variable and the independent variable, the unsafe behavior index is constructed.

5. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 3 is characterized in that: The grey relational analysis based on GRA includes; Identify, refine and categorize the specific causes of unsafe incidents based on the pilot's direct causes, indirect causes, root causes and root causes, and according to the detailed descriptions in the unsafe incident report; Based on the unsafe behavior indicators, the secondary influencing factors and their detailed reasons at four levels, namely, unsafe behavior, unsafe behavior premise, unsafe supervision and organizational influence, are counted using the 1 / 0 system; Calculate the grey correlation degree for the statistical results; According to the results of grey correlation calculation, the first-level, second-level and third-level indicators are analyzed respectively to analyze the deep-seated important factors that lead to the occurrence of unsafe behaviors; the first-level indicator is unsafe behavior, the second-level indicator is errors and violations, and the third-level indicators are skill errors, decision-making errors, perceptual errors, habitual violations and accidental violations; On the basis of grey correlation analysis, association rules are used to calculate and screen the support, confidence and improvement of ten influencing factors at three levels, namely, the premise of unsafe behavior, unsafe supervision and organizational influence, which lead to the occurrence of unsafe behavior; among them, the ten influencing factors include: environmental factors, operator status, personnel factors, insufficient supervision, inappropriate operation plan, uncorrected problems, supervision violations, resource management, organizational atmosphere and organizational process.

6. The method for quantitative mining of multi-level human factor associations of unsafe events according to claim 5 is characterized in that: Grey relational calculation includes: Determine the reference sequence and comparison sequence; the independent variables are the three levels of unsafe behavior premise, unsafe supervision and organizational influence and their underlying influencing factors, that is; the dependent variable is the unsafe behavior and its underlying influencing factors, that is, y′ j (r); Standardization processing: taking time as the third dimension to calculate the grey relational analysis: Where n represents the number of time periods divided according to the event distribution; r represents the time period; i and j represent the factors in the independent variable and the dependent variable respectively; xi(r) and yj(r) represent the original values ​​of the independent variable and the dependent variable respectively, and x'i(r) and y'j(r) represent the standardized values ​​of the independent variable and the dependent variable respectively; Compute the absolute difference of the sequence: D ij(r)＝ x′ i (r)-y′ j (r) D max (r)=maxΔ ij (r) D min (r)=minΔ ij (r) Among them, Δ ij (r) represents the difference between the independent variable and the dependent variable; Δ max (r) represents the maximum value among the differences; Δ min (r) represents the minimum value among the differences; Calculate the correlation coefficient: Among them, λij(r) represents the correlation coefficient; ρ represents the resolution coefficient, which is 0.5; (5) Calculate the degree of association; Among them, γij represents the correlation degree.

7. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 1 is characterized in that: The analysis results of the HFACS model are used to conduct a safety system vulnerability analysis based on the NK model for the eleven elements, including factors, software, hardware, cockpit environment, airport terrain environment, company operating environment, human and social environment, meteorological environment, airspace traffic environment, crew, and other people; The vulnerability analysis of security systems based on the NK model includes: Determine the number of risk elements N and the number of interdependent elements K; The analysis results of the HFACS model are used to conduct single, double and multi-factor vulnerability coupling analysis, and the NK model is used to conduct single, double and multi-factor vulnerability coupling analysis of risk factors in the subway construction safety system; Calculate the risk coupling probability and risk value, and evaluate the impact of different coupling factors on the total risk based on the risk occurrence probability and risk value under different coupling conditions; Combined with the vulnerability theory, the degree of factor coupling is used to measure the size of the system vulnerability.

8. The method for quantitative mining of multi-level human factor association of unsafe events according to claim 1 is characterized in that: The method further includes: performing a Bayesian network analysis based on BN on the dependency relationship result of the NK model; including: Identify risk events that need to be predicted and evaluated, and identify all factors that affect risk events; According to the dependencies between risk factors, the structure of Bayesian network is constructed; Use Bayes' theorem to calculate the posterior probability of risk events given evidence; Interpret the results of the Bayesian network analysis, including the relevance and importance of risk factors.

Images