Single pilot driving mode human factor reliability automatic prediction system and method

Abstract

The invention relates to the technical field of human factor reliability analysis, in particular to a system and a method for automatically predicting human factor reliability in a single pilot driving mode, and the system comprises a prediction unit I and a prediction unit II, the first prediction unit comprises a CPC evaluation module based on PIFs, a CPC membership degree module, a probability distribution reasoning module of a control mode and a human error probability evaluation module. The second prediction unit comprises a construction module, a failure mode recognition module, a second evaluation module and a human error probability calculation module. The method comprises the following steps: establishing an improved common performance condition table, and calculating a membership degree; constructing a Bayesian network; obtaining probability distribution of a plurality of control modes, and obtaining a human error probability value; establishing influence levels of the three added factors on the cognitive function, and performing cognitive function weight quantification according to the influence levels; a plurality of failure modes are newly added, and the basic fault probability CFPi of each failure mode is acquired; calculating a final human error probability; and obtaining the error probability of the total event.

Description

Technical Field

[0001] The present invention relates to the technical field of human reliability analysis, and in particular to a system and method for automatically predicting the human reliability of a single pilot's driving mode. Background Art

[0002] To reduce airline operating costs and mitigate safety risks associated with inconsistent cognition and operational controls in two-person crews, countries are conducting research on single-pilot flight models, a core technology for the development of next-generation commercial aircraft. In a two-person crew, when an emergency situation requires a single pilot, the addition of onboard decision-making assistance devices to the cockpit can offset the workload of the second person, redistributing the co-pilot's functions and responsibilities between the pilot and the automated systems, thereby increasing safety redundancy. However, increased cockpit automation also introduces new risks. If the pilot cannot accurately grasp the system's operating status, the workload increases and the single pilot's situational awareness is weakened. Furthermore, the addition of intelligent systems to the cockpit, while assisting the pilot in flight control, also increases the complexity of the human-machine interface, potentially leading to errors due to reliance on automation, lack of trust, and a lack of role recognition. However, due to human instability, human error is inevitable, so conducting human reliability analysis is an effective means of preventing errors and reducing accident risks.

[0003] Therefore, the present invention improves the automatic prediction method and system for human reliability based on the CREAM model, and reconstructs the common performance condition (CPC) indicator system for a single pilot's driving mode in combination with relevant research and expert opinions. Then, a precise calculation process is added to the basic method to convert the human error probability interval into an accurate numerical value; adaptive adjustments are made to the extended method for a single pilot's driving mode, adjusting possible failure modes and their basic error probabilities, and re-determining cognitive function weights; thus, an automated calculation system for predicting the pilot's human error probability is realized. Summary of the Invention

[0004] The present invention aims to provide a technical solution that adds a precise numerical calculation process to the basic method of the CREAM method and adjusts the applicability of the extended method to make it better suitable for the specific situation of a single pilot operation mode, thereby solving the problems in the background technology.

[0005] The technical solution of the present invention is: an automatic prediction system for human factor reliability of a single pilot's driving mode, characterized by comprising: a prediction unit 1 and a prediction unit 2;

[0006] The prediction unit 1 includes a CPC evaluation module based on PIFs, a CPC membership module, a control mode probability distribution reasoning module and a human error probability assessment module;

[0007] The PIFs-based CPC evaluation module is used to calculate the score of the first-level indicator CPC; the CPC membership module is used to obtain the membership of CPC; the reasoning module is used to obtain the probability distribution of multiple control modes; and the human error probability assessment module is used to calculate the probability of human error of the flight crew based on the probability distribution.

[0008] The second prediction unit includes a construction module, a failure mode identification module, a second evaluation module, and a human error probability calculation module;

[0009] The building blocks are used to automatically provide cognitive functions for different task sequences;

[0010] The failure mode identification module is used to display the basic error probability value;

[0011] The second evaluation module is a common performance condition evaluation module, which is used to select and input CPC and the second-level indicator behavior formation factor, automatically display the corresponding cognitive function weight factor, and calculate the total weight factor of the four types of cognitive functions respectively;

[0012] The human error probability calculation module is used to calculate the human error probability correction value of each task sequence and provide the final human error probability value in combination with event logic.

[0013] The present invention also provides a method for automatically predicting human reliability of a single pilot's driving mode, characterized in that the method comprises the following steps:

[0014] Establish an improved common performance condition table, establish the improved common performance condition and the membership function of the control mode, and calculate the membership degree;

[0015] The improved common performance conditions are: establishing common performance conditions that adapt to the single pilot's flying mode, specifically: deleting the original "team member cooperation quality" in the CREAM method and adding three factors: "ground-air communication quality", "pilot's own state", and "completeness of human-machine interface and intelligent assistance system";

[0016] Construct a Bayesian network and use it to infer the probability distribution of control modes to calculate the probability of human error;

[0017] Obtain the probability distribution of several control modes and use the weighted average method to defuzzify them to obtain the probability value of human error;

[0018] Establish the impact level of the three added factors on cognitive function, and quantify the weight of cognitive function according to the impact level;

[0019] Add several new failure modes and obtain the basic failure probability CFP0 of each failure mode;

[0020] Calculate the probability of human error for each failure mode CFPi = CFP0 × total weight factor;

[0021] Get the failure probability of the total event: Where n≥6.

[0022] Preferably, the membership function of the improved CPC and the control mode is established and the membership degree is calculated as follows: a trapezoidal membership function and a triangular membership function are selected to establish the membership function of the improved common performance condition and the control mode, and the membership degree is calculated by bringing in the evaluation score of the common performance condition;

[0023] The control modes include strategic, tactical, opportunistic and chaotic, and each control mode corresponds to a different error probability interval; the structure of the Bayesian network includes an input layer, an output layer and three auxiliary nodes.

[0024] Preferably, the method of using a Bayesian network to infer the probability distribution of the control mode to calculate the probability of human error includes:

[0025] Determine the discrete functions and conditional probability tables of the three nodes based on the improved mapping relationship diagram of common performance conditions and control modes;

[0026] Bayesian network reasoning is used to derive the probability distribution of control modes to calculate the probability of human error.

[0027] Preferably, the determining of the discrete functions and conditional probability tables of the three nodes based on the improved mapping relationship diagram of the common performance conditions and the control modes comprises the following steps:

[0028] For each CPC factor, define a discrete function g i :X i →E(i=1,2,…,9);

[0029] Where E = {improvement, not significant, reduction} represents the impact effect determined by the CPC level;

[0030] Define the discrete function of the auxiliary node: h:E 3 →G,

[0031] Where G = {(0,0), (0,1), (0,2), (0,3), (1,0), (1,1), (1,2), (2,0), (2,1), (3,0)}, and so on to derive the discrete function and conditional probability table of the control mode node.

[0032] Preferably, the method of obtaining the probability distribution of several control modes and defuzzifying them using a weighted average method to obtain a human error probability value comprises the following steps:

[0033] The influence of CPC level is integrated accordingly to obtain the probability distribution of control mode;

[0034] Defuzzification using weighted average method:

[0035] The calculation formula is:

[0036] Among them, u is the exact value after defuzzification, i is the number of fuzzy subsets, k i is the weighting coefficient selected according to actual conditions, x i is the membership degree of the control mode;

[0037] The final expression of the probability of human error is: HEP = 10 u ;

[0038] Calculate the probability of failure of the total event; the probability of failure of the total event is:

[0039] Preferably, the influence levels of the three factors added on cognitive function include:

[0040] The impact of the three factors of "ground-air communication quality", "pilot's own state", and "perfection of human-machine interface and intelligent assistance system" on cognitive function is divided into three levels: strong, medium, and weak;

[0041] Ignore the “weaker” effects and set the weight factor of CPC with weaker effects on cognitive function to 1, that is, no correction is made to the final probability of human error;

[0042] Since CPC has different levels, when the evaluation level is “not significant”, the weight factor assigned is 1;

[0043] When the evaluation level is “improved”, it indicates that there is a reduction effect on the probability of human error, and according to the impact level, “medium” is assigned a value of 0.8 and “strong” is assigned a value of 0.5;

[0044] When the evaluation level is “reduced”, it indicates that there is an improvement effect on the probability of human error, and according to the impact level, “medium” is assigned a value of 2.0 and “strong” is assigned a value of 5.0;

[0045] The newly added failure modes are: based on the factors of the newly added common performance conditions, 6 failure modes are added:

[0046] Failure modes of incomplete and unclear information perception and information misunderstanding: pilots misunderstand the information provided by intelligent assistance systems, interpretation bias, decision-making conflicts, insufficient trust, and perception and execution failures caused by over-reliance on automation.

[0047] Preferably, the step of establishing the improved common performance condition table further comprises the following steps:

[0048] The specific description of each CPC is refined into a secondary indicator structure, CPC factors are decomposed into multiple behavior formation factors, and factor screening is performed based on questionnaire surveys;

[0049] If the common performance condition in the row of the improved common performance condition table affects the common performance condition in the column, enter 1; otherwise, enter 0.

[0050] Using Ucinet software to form an influence relationship network diagram and calculate the betweenness centrality, the top 20% influence relationships were selected as key relationships, and the CPC adjustment rules were formed to improve the evaluation results according to the adjustment rules during the evaluation process of the expansion method;

[0051] The relevance and importance weights were calculated using the decision laboratory analysis method and the fuzzy analytic hierarchy process, respectively;

[0052] Perform comprehensive weight calculation;

[0053] The calculation of the relevance and importance weights respectively includes:

[0054] Calculate the relevance weight:

[0055] The relationship between CPCs is expressed on a scale of 0-5, where 0 means no influence, 1 means weak influence, 2 means relatively weak influence, 3 means average influence, 4 means strong influence, and 5 means very strong influence. Construct the direct influence matrix Z between CPCs k :

[0056]

[0057] Where k = 1, 2, ..., m, Z k is the direct influence matrix obtained by the k-th expert’s scoring, z ij k is the influence of the i-th indicator on the j-th indicator;

[0058] By normalizing the original relationship matrix, we get the standard matrix A k :

[0059]

[0060] Where,

[0061] Calculate the comprehensive influence matrix T between CPCs k :

[0062]

[0063] The CPC influence degree is obtained by adding the elements of the comprehensive influence matrix in rows and columns. and degree of influence

[0064] Calculate CPC centrality and causal degree The centrality is weighted by experts and then normalized to obtain the final correlation weight w i :

[0065]

[0066] Where λ k is the weight of the k-th expert;

[0067] Calculate importance weights:

[0068] Construct the importance judgment matrix,

[0069] Where B k is the importance judgment matrix given by the k-th expert, is the importance of the i-th indicator relative to the j-th indicator;

[0070] Convert the importance judgment matrix into the fuzzy consistent judgment matrix R k :

[0071]

[0072] Where,

[0073] The fuzzy consensus judgment matrices corresponding to multiple experts are weighted and aggregated to obtain the comprehensive fuzzy consistency matrix R:

[0074]

[0075] Where, λ k is the weight of the kth expert; calculate the row sum of the comprehensive fuzzy consistency matrix R,

[0076] Then perform normalization to obtain the importance weight:

[0077]

[0078] The comprehensive weight calculation is:

[0079] The weights obtained by DEMATEL and FAHP are linearly combined through combined weighting, and the adjustment parameter ρ is introduced with a value of 0.5. The comprehensive weight calculation formula is as follows: W = ρw + (1-ρ)w′;

[0080] Based on the experts' scores for each secondary indicator and the weights of the secondary indicators, the CPC score of the primary indicator is calculated using the weighted summation formula, and the evaluation is scored on a percentage basis. The specific formula is:

[0081] Where: m i is the number of corresponding secondary indicators, C i is the evaluation score of the i-th CPC, s ij is the secondary indicator score, w j is the corresponding secondary indicator weight.

[0082] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the method for automatically predicting human reliability of a single pilot driving mode.

[0083] Beneficial effects of the present invention:

[0084] 1. This invention understands the impact of increased automation levels and cockpit intelligent systems on pilots in single-pilot driving mode, analyzes factors influencing human reliability, and predicts the probability of human error. This helps determine the scenarios and cognitive stages in which pilots are most likely to make errors, thereby improving pilot working conditions and cockpit human-machine interface design in single-pilot driving mode. It also helps airlines and related organizations develop more reasonable operating standards and corresponding safety management policies, provides better training and support for the development of future pilots, and takes preventive measures against high-risk errors in advance, thereby improving aviation safety.

[0085] 2. The automatic human reliability prediction system provided by this invention helps improve calculation accuracy, save time and manpower, and can combine expert analysis and evaluation to predict the probability of human error for a single pilot's driving mode and provide the error rate of each cognitive function and the priority of improvement measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 This is a system structure diagram of the present invention;

[0087] Figure 2 Flow chart of the method of the present invention;

[0088] Figure 3 This is the Bayesian network structure diagram of the present invention;

[0089] Figure 4 Schematic diagram of control mode probability distribution of the present invention;

[0090] Figure 5 is a flow chart of a module of a prediction unit of the present invention;

[0091] Figure 6 It is the overall logical relationship diagram of the method of the present invention. DETAILED DESCRIPTION

[0092] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are for illustrative purposes only, and those skilled in the art will readily appreciate other obvious variations. The basic principles of the present invention defined in the following description may be applied to other embodiments, variations, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the present invention.

[0093] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the elements may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0094] refer to Figure 1 ,The present invention provides a technical solution: an automatic prediction system for human reliability of a single pilot's driving mode, comprising: a prediction unit 1 and a prediction unit 2;

[0095] The prediction unit 1 includes a CPC evaluation module based on PIFs, a CPC membership module, a control mode probability distribution reasoning module and a human error probability assessment module; the prediction unit 2 includes a construction module, a failure mode identification module, an evaluation module 2 and a human error probability calculation module; PIFs are the expert scores for the second-level indicator behavior formation factor (PIF).

[0096] refer to Figure 5 , the functions implemented by the prediction unit 1 are as follows:

[0097] According to the weights of PIFs and secondary indicators, the score of the primary indicator CPC is calculated using the weighted summation formula, and the percentage system is used for evaluation.

[0098] The CPC scores obtained in the PIFs-based CPC evaluation submodule are input into the membership function of the common performance condition to obtain the membership of each CPC.

[0099] The membership of CPC is input into the Bayesian network, and the probability distribution of the four control modes is inferred by combining the set discrete function and conditional probability table.

[0100] The final pilot error probability is calculated based on the probability distribution of the control mode using the weighted average defuzzification method.

[0101] The probability of failure of the total event described is:

[0102] The functions implemented by prediction unit 2 are as follows:

[0103] Decompose the pilot's tasks, input the task sequence and its corresponding cognitive activities, and automatically give the cognitive functions of different task sequences based on the correspondence between cognitive activities and cognitive functions.

[0104] Based on the cognitive function given in the previous module, the most likely failure mode is input and the corresponding basic failure probability value is automatically displayed.

[0105] According to expert opinions, the levels of CPC and PIF are selected and input, the corresponding cognitive function weight factors are automatically displayed, and the total weight factors of the four types of cognitive functions are calculated separately.

[0106] The human error probability correction value of each task sequence is automatically calculated according to the calculation formula, and the final human error probability value is given in combination with event logic.

[0107] refer to Figure 2 、 Figure 6 The present invention also provides a method for automatically predicting the human reliability of a single pilot's driving mode:

[0108] Establish an improved common performance condition table, establish membership functions for the improved common performance conditions and control modes, and calculate the membership degree. Specifically, select the domain of the CPC factor fuzzy set as [0, 100], select trapezoidal membership functions and triangular membership functions, establish membership functions for the common performance conditions and control modes, and substitute the evaluation scores of the common performance conditions into the membership functions to calculate the membership degree.

[0109] The CREAM method provides four types of control modes: strategic, tactical, opportunistic, and chaotic. Each control mode corresponds to a different error probability range. The specific ranges are shown in Table 1.

[0110] Table 1 Control modes and their error probability intervals

[0111] Control Mode Error probability interval Strategic <![CDATA[5×10 -5 ﹤P﹤1×10 -2 ]]> Tactical <![CDATA[1×10 -3 ﹤P﹤1×10 -1 ]]> Opportunistic <![CDATA[1×10 -2 ﹤P﹤5×10 -1 ]]> Chaotic <![CDATA[1×10 -1 ﹤P﹤1]]>

[0112] The improved common performance conditions are established to adapt to a single pilot's driving mode. Specifically, the original "team member cooperation quality" factor in the CREAM method is removed, and three factors are added: "ground-to-air communication quality," "pilot's own state," and "the completeness of the human-machine interface and intelligent assistance system." The improved common performance conditions are shown in Table 2:

[0113] Table 2 Improved common performance conditions

[0114]

[0115] like Figure 4, construct a Bayesian network, and use the Bayesian network to infer the probability distribution of the control mode. To calculate the probability of human error; specifically:

[0116] The membership of the common performance condition is used as input and the probability distribution of the control mode is used as output. At the same time, in order to simplify the calculation and reduce the load of the software, three auxiliary nodes are introduced between the input layer and the output layer to establish a Bayesian network diagram, such as Figure 3 Based on the improved mapping relationship diagram between common performance conditions and control modes, the discrete functions and conditional probability tables of the three-level nodes are determined, and the probability distribution of the control mode is obtained using Bayesian network reasoning to calculate the probability of human error.

[0117] For each CPC factor, define a discrete function g i :X i →E(i=1,2,…,9); where E={improvement, not significant, reduction} represents the expected effect determined by the CPC level.

[0118] Define the discrete function h:E of the auxiliary node 3 →G;

[0119] Where G = {(0,0), (0,1), (0,2), (0,3), (1,0), (1,1), (1,2), (2,0), (2,1), (3,0)}, and so on to derive the discrete function and conditional probability table of the control mode node.

[0120] The probability distribution of several control modes is obtained and defuzzified using the weighted average method to obtain the probability value of human error. The impact level of the three added factors on cognitive function is established and the cognitive function weight is quantified according to the impact level. Specifically:

[0121] Select the weighted average method for defuzzification, and the calculation formula is as follows:

[0122] In the formula, u is the exact value after defuzzification, i is the number of fuzzy subsets, k is the i is the weighting coefficient selected according to actual conditions, x i is the degree of membership of the control mode. Since the logarithm of the value interval of the control mode is taken before fuzzification, the final expression of the human error probability (HEP) is: HEP = 10 u .

[0123] The implementation method of the extension method in CREAM in this embodiment is as follows:

[0124] Since the present invention identifies new common performance conditions for a single pilot's flying mode, the weights of CPC corresponding to cognitive functions in the original CREAM method can no longer be used. Therefore, empirical data and expert judgment are used to reanalyze the basic relationship between the newly added CPC factor and the four types of cognitive functions, and the impact level is quantified into cognitive function weights corresponding to multiple levels.

[0125] Because the completeness of the human-machine interface and intelligent assistance system, the pilot's own status, and the quality of ground-to-air communication are newly proposed common performance conditions for the SPO model, only these three items are analyzed, and their impact on cognitive function is divided into three levels: strong, medium, and weak. The resulting impact level table is shown in Table 3.

[0126] Table 3. Impact of the newly added common performance conditions on cognitive function

[0127]

[0128]

[0129] The rules for quantifying the weight of cognitive function according to the impact level are as follows: first, ignore the "weaker" impact and set the weight factor of CPC with weaker impact on cognitive function to 1, that is, no correction is made to the final probability of human error; second, since CPC has different levels, when the evaluation level is "not significant", the weight factor assigned is 1; when the evaluation level is "improved", it indicates that there is a reducing effect on the probability of human error, and according to the impact level of "medium", the value is assigned to 0.8, and "strong" is assigned to 0.5; when the evaluation level is "reduced", it indicates that there is an increasing effect on the probability of human error, and according to the impact level of "medium", the value is assigned to 2.0, and "strong" is assigned to 5.0.

[0130] The CREAM methodology categorizes human cognitive functions into four main categories: observation, interpretation, planning, and execution. Each category has several failure modes, for a total of 13 failure modes. However, to achieve single-pilot operation, the next-generation intelligent cockpit requires a higher level of automation, which differs from two-pilot flight crews and may introduce new failure modes.

[0131] A questionnaire survey was conducted to collect experts' opinions on the probability of occurrence of the newly added error modes, and the basic error probability values ​​of each failure mode were adjusted. Based on the improved original CREAM method, a human factors event analysis was conducted under the single pilot driving mode.

[0132] Add several new failure modes and obtain the basic failure probability CFP of each failure mode i ;

[0133] Specifically: Based on relevant research and expert opinions, the following failure modes are added:

[0134] (1) Incomplete and unclear information perception: In the intelligent cockpit environment, there are too many sources of information input, and the pilot does not perceive all the information or the perceived information is irrelevant;

[0135] (2) Information misunderstanding: pilots misunderstand the information provided by the intelligent assistance system;

[0136] (3) Interpretation bias: pilots do not understand the complex automation system or lack relevant experience, and are unable to judge the current situation, resulting in a deviation in their understanding of the situation;

[0137] (4) Decision conflict: When the operation prompts given by the intelligent assistance system are inconsistent with the pilot's decision, the pilot will be confused in the decision-making process;

[0138] (5) Lack of trust: When pilots do not fully understand the intelligent assistance system, they ignore the auxiliary information and blindly perform operations that they think are feasible.

[0139] (6) Perception and execution failure caused by over-reliance on automation: Pilots’ over-reliance on intelligent assistance systems may lead to excessive focus on the assistance systems and neglect of the surrounding environment, resulting in the degradation of manual skills during execution.

[0140] Calculate the final human error probability CFP i =CFP0×total weight factor;

[0141] Get the total event failure probability

[0142] The nine CPC factors used to describe the task scenario environment have different effects on human reliability; rather, they have certain dependencies. Therefore, social network analysis was used to analyze the relationship between factors. This method is primarily used to analyze the associations between individuals in a network, reflecting the structure and attributes of relationships. An influence relationship adjacency matrix was constructed using expert judgment. If the common performance condition in a row affects the common performance condition in a column, a 1 was entered; otherwise, a 0 was entered. Based on this, Ucinet software was used to form an influence relationship network diagram and calculate the betweenness centrality. The top 20% of influence relationships were selected as key relationships to form CPC adjustment rules. This adjustment rule was then used to improve the evaluation results during the evaluation process of the extended method.

[0143] The CPC weights were calculated using the Decision Analysis Laboratory (DEMATEL) method and the Fuzzy Analytic Hierarchy Process (FAHP) method to calculate the relevance and importance weights, respectively.

[0144] Correlation weight calculation: DEMATEL is a system analysis method that uses graph theory and matrix tools. It does not have a rigid requirement for the independence of the indicators being analyzed. By using the logical relationships and direct influence matrices between the indicators in the system, it can calculate the influence of each indicator on other indicators, as well as the degree of influence, causality, and centrality. It can effectively reflect the influence relationship between different indicators and is an important method for analyzing the correlation between factors. This study uses the DEMATEL method to calculate the correlation weight of CPC. The specific steps are as follows:

[0145] 1) Quantify the relationship between each indicator and obtain the direct impact matrix. The relationship between CPCs is expressed on a scale of 0-5, where 0 means no impact, 1 means weak impact, 2 means relatively weak impact, 3 means average impact, 4 means strong impact, and 5 means very strong impact. Construct the direct impact matrix Z between CPCs. k :

[0146] Where Z k The direct influence matrix obtained by scoring the k-th expert. is the influence of the i-th indicator on the j-th indicator, k = 1, 2, 3…, m.

[0147] 2) By normalizing the original relationship matrix, we get the standard matrix A k :

[0148] Where,

[0149] 3) Calculate the comprehensive influence matrix T between CPCs k :

[0150] 4) The influence degree of CPC is obtained by adding the elements of the comprehensive influence matrix in rows and columns respectively. and degree of influence

[0151] 5) Calculate CPC centrality and causal degree The centrality is weighted by experts and then normalized to obtain the final correlation weight w i :

[0152]

[0153] Where λ k is the weight of the k-th expert.

[0154] (2) Importance weight calculation

[0155] The fuzzy analytic hierarchy process (FAHP) improves the problems of the traditional analytic hierarchy process and enhances the reliability of decision making. Combining the advantages of fuzzy method and analytic hierarchy process, the fuzzy analytic hierarchy process (FAHP) can effectively solve the problem of difficulty in ensuring the consistency of thinking when there are too many evaluation indicators at a certain level in AHP. This study uses the fuzzy analytic hierarchy process to assign importance weights. The specific steps are as follows:

[0156] 1) Construct an importance judgment matrix: Compare and score the relative importance of each pair of factors in the same level and affiliation. A nine-level scale of 0.1-0.9 is used for dimension segmentation. The specific meaning is shown in Table 4 below:

[0157] Table 4 Meaning of the nine-level scale

[0158]

[0159] Importance judgment matrix:

[0160] Where B k is the importance judgment matrix given by the k-th expert, is the importance of the i-th indicator relative to the j-th indicator.

[0161] 2) Convert the importance judgment matrix into the fuzzy consistency judgment matrix R k :

[0162] Where,

[0163] 3) The fuzzy consensus judgment matrices corresponding to multiple experts are weighted and aggregated to obtain the comprehensive fuzzy consistency matrix R:

[0164] Where, λ k is the weight of the k-th expert.

[0165] 4) Calculate the importance weight: Calculate the row sum of the comprehensive fuzzy consistency matrix R, and then perform normalization to obtain the importance weight:

[0166] (3) Comprehensive weight calculation

[0167] In order to obtain a comprehensive weight that takes into account the importance of CPC itself and the correlation between CPCs, the weights obtained by DEMATEL and FAHP are linearly combined through combined weighting, and the adjustment parameter ρ is introduced with a value of 0.5. The comprehensive weight calculation formula is as follows: W = ρw + (1-ρ)w′;

[0168] Based on the experts' scores for each secondary indicator and the weights of the secondary indicators, the CPC score of the primary indicator is calculated using the weighted summation formula, and the evaluation is scored on a percentage basis. The specific formula is as follows: Where: m i is the number of corresponding secondary indicators, C i is the evaluation score of the i-th CPC, s ij is the secondary indicator score, w j is the corresponding secondary indicator weight.

[0169] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the method for automatically predicting human reliability of a single pilot driving mode.

[0170] The embodiments disclosed in the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. The embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part, and / or installed from a removable medium. When the computer program is executed by the central processing unit (CPU), the above-mentioned functions defined in the method of the present application are executed. It should be noted that the computer-readable medium mentioned above in the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or component of an electrical, magnetic, optical, electromagnetic, infrared segment, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wire segments, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, an electromagnetic signal, an optical signal, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, electrical wire, optical fiber cable, RF, etc., or any suitable combination thereof.

[0171] The flow charts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0172] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functional and structural principles of the present invention have been demonstrated and explained in the embodiments. Without departing from the principles, the implementation methods of the present invention may be subject to any deformation or modification.

Claims

1. An automatic prediction system for human reliability of a single pilot's driving mode, characterized in that: include: Prediction unit one and prediction unit two; The prediction unit 1 includes a CPC evaluation module based on PIFs, a CPC membership module, a control mode probability distribution reasoning module and a human error probability assessment module; The CPC evaluation module based on PIFs is used to calculate the score of the first-level indicator CPC according to the expert's score on the second-level indicator behavior formation factor; the CPC membership module is used to obtain the membership of CPC; the reasoning module is used to obtain the probability distribution of multiple control modes; the human error probability evaluation module is used to calculate the probability of human error of the flight personnel according to the probability distribution; The prediction unit 2 includes a construction module, a failure mode identification module, an evaluation module 2, and a human error probability calculation module; The building blocks are used to automatically provide cognitive functions for different task sequences; The failure mode identification module is used to display the basic error probability value; The evaluation module 2 is a common performance condition evaluation module, which is used to select and input CPC and the second-level indicator behavior formation factor, automatically display the corresponding cognitive function weight factor, and calculate the total weight factor of the four types of cognitive functions respectively; The human error probability calculation module is used to calculate the human error probability correction value of each task sequence and give the final human error probability value in combination with the event logic.

2. A method for automatically predicting the human factor reliability of a single pilot's driving mode, characterized in that: The method comprises the following steps: Establish an improved common performance condition table, establish the improved common performance condition and the membership function of the control mode, and calculate the membership degree; The improved common performance conditions are: establishing common performance conditions that adapt to the single pilot's driving mode, specifically: deleting the original "collaboration quality of team members" in the CREAM method, and adding three factors: "ground-air call quality", "pilot's own state", and "perfection of human-machine interface and intelligent assistance system"; Construct a Bayesian network and use it to infer the probability distribution of control modes to calculate the probability of human error; Obtain the probability distribution of several control modes and use the weighted average method to defuzzify them to obtain the probability value of human error; Establish the impact levels of the three added factors on cognitive function, and quantify the cognitive function weights according to the impact levels; Add several new failure modes and obtain the basic failure probability CFP0 of each failure mode; Calculate the human error probability value CFP for each failure mode i =CFP0×total weight factor; Calculate the probability of failure for the total event: Where n≥6.

3. The method for automatically predicting human factor reliability of a single pilot's driving mode according to claim 2, characterized in that: The membership function of the CPC and the control mode is calculated as follows: the trapezoidal membership function and the triangular membership function are selected to establish the improved common performance condition and the membership function of the control mode, and the membership is calculated by bringing in the evaluation score of the common performance condition; The control modes include strategic, tactical, opportunistic and chaotic, and each control mode corresponds to a different error probability interval; the structure of the Bayesian network includes an input layer, an output layer and three auxiliary nodes.

4. The method for automatically predicting human factor reliability of a single pilot's driving mode according to claim 3, characterized in that: The method of using the Bayesian network to infer the probability distribution of the control mode to calculate the probability of human error includes: Determine the discrete functions and conditional probability tables of the three nodes based on the improved mapping relationship diagram of the common performance conditions and the control mode; Bayesian network reasoning is used to derive the probability distribution of control modes to calculate the probability of human error.

5. The method for automatically predicting human factor reliability of a single pilot's driving mode according to claim 4, characterized in that: The method of determining the discrete functions and conditional probability tables of the three nodes based on the improved mapping relationship diagram of the common performance conditions and the control mode includes the following steps: For each CPC factor, define a discrete function g i :X i →E(i=1,2,…,9); Among them, E = {improvement, not significant, reduction} represents the impact effect determined by the CPC level; Define the discrete function of the auxiliary node: h:E 3 →G, Among them, G={(0,0),(0,1),(0,2),(0,3),(1,0),(1,1),(1,2),(2,0),(2,1),(3,0)}, and so on to derive the discrete function and conditional probability table of the control mode node.

6. The method for automatically predicting human factor reliability of a single pilot's driving mode according to claim 4, characterized in that: The method of obtaining the probability distribution of several control modes and defuzzifying them by weighted average method to obtain the probability value of human error comprises the following steps: The influence of CPC level is integrated accordingly to obtain the probability distribution of control mode; Use weighted average method to defuzzify: The calculation formula is: Among them, u is the exact value after defuzzification, i is the number of fuzzy subsets, k i is the weighting coefficient selected according to the actual situation, x i is the degree of membership of the control mode; The final expression of the probability of human error is: HEP = 10 u .

7. The method for automatically predicting human reliability of a single pilot's driving mode according to claim 2, characterized in that: The three factors that were established to increase the level of influence on cognitive function include: The influence of the three factors of "ground-air call quality", "pilot's own state" and "perfection of human-machine interface and intelligent assistance system" on cognitive function is divided into three levels: strong, medium and weak; Ignore the "weaker" effects and set the weight factor of CPC with weaker effects on cognitive function to 1, that is, no correction is made to the final probability of human error; Since CPC has different levels, when the evaluation level is "not significant", the weight factor assigned is 1; When the evaluation level is "improved", it means that the probability of human error is reduced. According to the impact level, "medium" is assigned a value of 0.8, and "strong" is assigned a value of 0.5; When the evaluation level is "reduced", it means that the probability of human error is improved. According to the impact level, "medium" is assigned a value of 2.0 and "strong" is assigned a value of 5.

0.

8. The method for automatically predicting human reliability of a single pilot's driving mode according to claim 2, characterized in that: The newly added failure modes are: based on the factors of the newly added common performance conditions, 6 failure modes are added: Failure modes of incomplete and unclear information perception and information misunderstanding: pilots misunderstand the information provided by the intelligent assistance system, interpretation bias, decision conflict, lack of trust, and perception and execution failures caused by over-reliance on automation.

9. The method for automatically predicting human reliability of a single pilot's driving mode according to claim 2, characterized in that: The said establishing the improved common performance condition table further comprises the following steps: The specific description of each CPC is refined into a secondary indicator structure, CPC factors are decomposed into multiple behavior formation factors, and factor screening is performed based on questionnaire surveys; If the common performance condition of the row in the improved common performance condition table will affect the common performance condition of the column, enter 1, otherwise enter 0; The influence relationship network diagram was formed using Ucinet software and the betweenness centrality was calculated. The top 20% influence relationships were selected as key relationships and the CPC adjustment rules were formed to improve the evaluation results according to the adjustment rules in the evaluation process of the extended method. The relevance and importance weights were calculated by decision laboratory analysis and fuzzy analytic hierarchy process, respectively; Perform comprehensive weight calculation; The respectively calculating the relevance and importance weight comprises: Calculate the relevance weight: The relationship between CPCs was expressed using a 0-5 scale, where 0 means no effect, 1 means a weak effect, 2 means a relatively weak effect, 3 means a moderate effect, 4 means a relatively strong effect, and 5 means a very strong effect; Construct the direct influence matrix Z between CPCs k : In the formula, Z k (k=1,2,...,m) is the direct impact matrix obtained by the kth expert’s scoring, z ij k is the influence of the i-th indicator on the j-th indicator; By normalizing the original relationship matrix, we get the standard matrix A k : In the formula, Calculate the comprehensive influence matrix T between CPCs k : The influence of CPC is obtained by adding the elements of the comprehensive influence matrix in rows and columns. and the degree of influence F i k : Calculate CPC centrality and cause degree After the centrality is weighted by experts and then normalized, the final correlation weight w is obtained. i : In the formula, λ k is the weight of the kth expert; Calculate importance weights: Construct the importance judgment matrix, In the formula, B k is the importance judgment matrix given by the k-th expert, is the importance of the i-th indicator relative to the j-th indicator; Convert the importance judgment matrix into the fuzzy consistent judgment matrix R k : In the formula, The fuzzy consistent judgment matrices corresponding to multiple experts are weighted and aggregated to obtain the comprehensive fuzzy consistency matrix R: In the formula, λ k is the weight of the kth expert; the row sum of the comprehensive fuzzy consistency matrix R is calculated and then normalized to obtain the importance weight: The comprehensive weight calculation is: The weights obtained by DEMATEL and FAHP are linearly combined through combined weighting, and the adjustment parameter ρ is introduced with a value of 0.

5. The comprehensive weight calculation formula is as follows: W = ρw + (1-ρ)w′; According to the experts' scores for each secondary indicator and the weight of the secondary indicators, the CPC score of the primary indicator is calculated using the weighted summation formula, and the evaluation is based on a percentage system. The specific formula is as follows: Where: m i is the number of corresponding secondary indicators, C i is the evaluation score of the ith CPC, s ij is the secondary indicator score, w j is the corresponding secondary indicator weight.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method for automatically predicting the human reliability of a single pilot driving mode as described in any one of claims 2 to 9.

Images