Civil aviation maintenance risk prediction framework construction method and system

Abstract

The invention discloses a civil aviation maintenance risk prediction framework construction method and system, and belongs to the technical field of civil aviation maintenance risk evaluation and management. The method comprises the following steps: S1, collecting multi-source maintenance data and a maintenance manual standard to obtain a multi-source data set; s2, performing event and risk variable classification based on the multi-source data set to obtain a Bayesian network variable; constructing a Bayesian maintenance risk network based on the Bayesian network variable, wherein the Bayesian maintenance risk network is used for risk analysis; and S3, constructing a dynamic evolution model based on a risk analysis result, and performing decision analysis based on the dynamic evolution model. According to the method, the Bayesian network and the system dynamics are adopted to carry out risk analysis on the maintenance scene and carry out organization level decision evolution, so that the risk probability in the civil aviation maintenance scene is expected to be evaluated, and the risk development situation under different behavior decisions is explored.

Description

Technical Field

[0001] The present invention belongs to the technical field of civil aviation maintenance risk assessment and management, and in particular relates to a method and system for constructing a civil aviation maintenance risk prediction framework. Background Art

[0002] Civil aviation authorities have long prioritized the maintenance quality management system for aircraft and aircraft components as a key safety management focus, with clear regulations establishing this requirement. The Quality and Safety Management System for Maintenance Organizations (AC-145-FS-015R1) requires that maintenance quality risks be measured.

[0003] As the digitalization process of the aviation maintenance industry gradually accelerates, big data technology is used to analyze and model the large amount of data generated during the daily operation, maintenance and repair of aircraft. The maintenance method has changed from the original "diagnosis-based planned maintenance" to the current "predictive maintenance based on mathematical statistics."

[0004] Traditional maintenance quality management is insufficiently reliable. Maintenance personnel must spend considerable time studying troubleshooting manuals, maintenance operating procedures, and failure analysis reports to understand the underlying risk sources and their interdependencies, placing a significant burden on risk analysis. Risk management only addresses maintenance activities and mechanical failures, proposing risk mitigation solutions and lacking research on the evolution of maintenance organizations. Summary of the Invention

[0005] The present invention aims to address the deficiencies of the prior art and proposes a method for constructing a civil aviation maintenance risk prediction framework, which can better evaluate maintenance operation risks.

[0006] To achieve the above objectives, the present invention provides the following solution: a method for constructing a civil aviation maintenance risk prediction framework, comprising the following steps:

[0007] S1, collect multi-source maintenance data and maintenance manual standards to obtain a multi-source data set;

[0008] S2. Classifying events and risk variables based on the multi-source data set to obtain Bayesian network variables; and constructing a Bayesian maintenance risk network based on the Bayesian network variables, wherein the Bayesian maintenance risk network is used for risk analysis;

[0009] S3. Based on the results of the risk analysis, a dynamic evolution model is constructed, and a decision analysis is performed based on the dynamic evolution model.

[0010] Further preferably, the Bayesian network variables include: events, environmental factors, installation and commissioning errors, maintenance procedures and management defects, parts and tool management omissions, personnel and training defects, documentation and compliance issues, supervision and resource issues, time and technical pressures.

[0011] Further preferably, the method of constructing the Bayesian maintenance risk network based on the Bayesian network variables in S2 includes: identifying dependency relationships of the Bayesian network variables based on structure learning, and constructing the Bayesian maintenance risk network based on the dependency relationships;

[0012] Including: S21, for each node of the Bayesian maintenance risk network, obtain the joint probability distribution based on parameter learning:

[0013]

[0014] Where: P(R1,…,R n , C) represents the joint probability distribution; the severity C of the maintenance event is the risk variable set R={R1,...,R i} is the only parent variable; P(C) represents the probability of maintenance events; n represents the number of risk variables; P(R i |C) represents the probability of risk variables under maintenance events;

[0015] S23. Introduce a tree structure for risk variables, add additional edges from data learning to represent the dependencies between risks, and obtain the functional relationship:

[0016]

[0017] In the formula, {C,R π(i)} represents the relationship between maintenance event C and a risk variable R i The parent node R π(i) The relationship between them; {C} represents the maintenance event; π(i) represents the function π that converts each risk variable R i Mapped to its parent node.

[0018] S25. Based on the joint probability distribution and the functional relationship, obtain the posterior probability of the severity of the maintenance event:

[0019]

[0020] Where, P(C|R1,…,R n ) represents the known risk variables R1,...,R n The probability of maintenance event C occurring under the conditions of their respective specific states; P(R root |C) represents the root node risk variable R root Conditional probability under a given class variable C state; P(R i |C,∏R i ) represents the risk variable R i Given a class variable C and its parent node set ΠR i Conditional probability under state;

[0021] S27. Based on the given severity of the maintenance event, the dependency relationship between the two risk variables is obtained:

[0022]

[0023] Where, I P represents the conditional mutual information between two risk variables; r ii Represents the risk variable R i The i-th state; r ji Represents the risk variable R j The i-th state; c i Indicates the i-th state of the severity of the maintenance event; b is a constant; P(r ii ,r ji |c i ) means that given C=c i Under the condition of i and R j The joint conditional probability of P(r ii |c i ) means that given C=c i Under the condition of i The marginal conditional probability of P(r ji |c i ) means that given C=c i Under the condition of j The marginal conditional probability of .

[0024] Further preferably, the method for performing risk analysis based on the Bayesian maintenance risk network in S2 includes:

[0025] S22. Calculate the probability of occurrence of accident risk factors for the Bayesian maintenance risk network using the software GeNIe, and update the probability state of the Bayesian maintenance risk network without providing evidence or changing any node state;

[0026] S24, by setting the state of the event risk variable to State1 in the GeNIe software, observing the probability change of the parent variable, and updating the Bayesian maintenance risk network to obtain the posterior probability of the parent variable of the risk variable;

[0027] S26. Setting the risk variable as the target node in the Bayesian maintenance risk network to perform sensitivity analysis.

[0028] Further preferably, based on the results of the risk analysis, the maintenance error factors are classified into: technical errors and non-technical errors;

[0029] For technical errors, make corrections based on the maintenance manual;

[0030] For non-technical errors, the dynamic evolution model is constructed, and decision analysis is performed based on the dynamic evolution model.

[0031] Further preferably, the maintenance fault is represented as:

[0032] F(t)=F(t-dt)+R F (t)-fault rectification rate (t)*dt,

[0033] R F (t)=k7*(1-P(t))*(1-S(t)*(1-C(t)),

[0034] Where F(t) represents the number of maintenance failures; F(t-dt) represents the historical cumulative number of maintenance failures at time point t-dt; R F (t) represents the failure rate; k7 represents the influence coefficient of the combined effects of the maintenance procedure perfection P(t), personnel skill level S(t), and inspection adequacy C(t); P(t) represents the maintenance procedure perfection; S(t) represents the personnel skill level.

[0035] Further preferably, the causal loop of the dynamic evolution model includes four feedback loops:

[0036] P(t)=P(t-dt)+R p (t)*dt,

[0037] R p (t)=k3*(P 目标 -P(t))+k4*A(t),

[0038] Where P(t) represents the degree of perfection of the maintenance program; P(t-dt) represents the historical cumulative degree of perfection of the maintenance program at time point t-dt; R p (t) represents the program improvement rate; k3 represents the influence coefficient of the maintenance program perfection degree; k4 represents the influence coefficient of the program information accuracy; P 目标 It represents the expected degree of perfection of the maintenance program; A(t) represents the accuracy of the program information;

[0039] S(t)=S(t-dt)+R S (t)*dt,

[0040] R s (t) = k1*(S 目标 -S(t))+k2*G(t),

[0041] In the formula, S(t) represents the skill level of personnel; S(t-dt) represents the historical cumulative skill level of personnel at the time point t-dt; R S (t) represents the training rate; k1 represents the influence coefficient of personnel skill level; k2 represents the influence coefficient of communication effect; S 目标 It represents the expected skill level of personnel; G(t) represents the communication effect;

[0042] M(t)=M(t-dt)+R M (t)*dt,

[0043] R M (t)=k5*(M 目标 -M(t))+k6*F(t),

[0044] Where M(t) represents the supervision intensity; M(t-dt) represents the historical cumulative supervision intensity at time point t-dt; M 目标 Indicates the expected supervision intensity; R M (t) represents the supervision strengthening rate; k5 represents the supervision intensity influence coefficient, and k6 represents the maintenance failure number influence coefficient;

[0045] And auxiliary variables: process information accuracy A(t), inspection adequacy C(t), communication effectiveness G(t):

[0046] A(t)=P(t)*k8,

[0047] C(t)=C(t)*k9,

[0048] G(t)=S(t)*k 10 ,

[0049] Where, k8 represents the influence coefficient of the degree of perfection of maintenance procedures on the accuracy of procedure information; k9 represents the influence coefficient of supervision intensity on the adequacy of inspection; k 10 It represents the influence coefficient of factors such as personnel training intensity and organizational management on the accuracy of communication effectiveness.

[0050] The present invention also provides a civil aviation maintenance risk prediction framework construction system, comprising:

[0051] A data acquisition module is used to collect multi-source maintenance data and maintenance manual standards to obtain a multi-source data set;

[0052] A model building module is used to classify events and risk variables based on the multi-source data set to obtain Bayesian network variables; and to build a Bayesian maintenance risk network based on the Bayesian network variables, wherein the Bayesian maintenance risk network is used for risk analysis;

[0053] The decision analysis module is used to construct a dynamic evolution model based on the results of the risk analysis and perform decision analysis based on the dynamic evolution model.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] This paper adopts Bayesian networks and system dynamics to conduct risk analysis on maintenance scenarios and conduct decision evolution at the organizational level, in order to evaluate the risk probability in civil aviation maintenance scenarios and explore the risk development trend under different behavioral decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. 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 any creative work.

[0057] Figure 1 This is a flowchart of a method for constructing a civil aviation maintenance risk prediction framework according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0059] 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.

[0060] Example 1:

[0061] like Figure 1 As shown, this embodiment provides a method for constructing a civil aviation maintenance risk prediction framework, including the following steps:

[0062] S1. Collect multi-source maintenance data and maintenance manual standards to obtain a multi-source dataset.

[0063] S2. Classify events and risk variables based on multi-source data sets to obtain Bayesian network variables; and construct a Bayesian maintenance risk network based on the Bayesian network variables. The Bayesian maintenance risk network is used for risk analysis.

[0064] In this embodiment, the Bayesian network variables include: events, environmental factors, installation and commissioning errors, maintenance procedures and management defects, parts and tool management omissions, personnel and training defects, documentation and compliance issues, supervision and resource issues, time and technical pressure.

[0065] The method for constructing a Bayesian maintenance risk network based on Bayesian network variables includes: identifying dependency relationships of Bayesian network variables based on structure learning, and constructing a Bayesian maintenance risk network based on the dependency relationships.

[0066] Specifically including: S21, for each node of the Bayesian maintenance risk network, the joint probability distribution describing its parent node is obtained based on parameter learning:

[0067]

[0068] Where: P(R1,…,R n , C) represents the joint probability distribution; the severity C of the maintenance event is the risk variable set R={R1,...,R i} (i.e., Π={C}, 1≤i≤10), and C has no parent variable (i.e. ); P(C) represents the probability of maintenance events; n represents the number of risk variables; P(R i |C) represents the probability of risk variables under maintenance events.

[0069] S23. Introduce a tree structure for risk variables, add additional edges from data learning to represent the dependencies between risks, and obtain the functional relationship:

[0070]

[0071] In the formula, {C,R π(i)} represents the relationship between maintenance event C and a risk variable R i The parent node R π(i) The relationship between them; {C} represents the maintenance event; π(i) represents the function π that converts each risk variable R i Mapped to its parent node.

[0072] S25. Based on the joint probability distribution and the functional relationship, the posterior probability of the severity of the maintenance event is obtained:

[0073]

[0074] Where, P(C|R1,…,R n ) represents the known risk variables R1,...,R n The probability of maintenance event C occurring under the conditions of their respective specific states; P(R root |C) represents the root node risk variable R rootConditional probability under a given class variable C state; P(R i |C,∏R i ) represents the risk variable R i Given a class variable C and its parent node set ∏R i The conditional probability of the state.

[0075] S27. Based on the given maintenance event severity, the dependency relationship between the two risk variables is obtained. This value is used as the weight of the arc to obtain a tree with the maximum sum of the weights of all arcs:

[0076]

[0077] Where, I P represents the conditional mutual information between two risk variables; r ii Represents the risk variable R i The i-th state; r ji Represents the risk variable R j The i-th state; c i Indicates the i-th state of the severity of the maintenance event; b is a constant; P(r ii ,r ji |c i ) means that given C=c i Under the condition of i and R j The joint conditional probability of P(r ii |c i ) means that given C=c i Under the condition of i The marginal conditional probability of P(r ji |c i ) means that given C=c i Under the condition of j The marginal conditional probability of .

[0078] The methods for risk analysis based on Bayesian maintenance risk network include:

[0079] S22. Using the software GeNIe, the probability of occurrence of accident risk factors is calculated for the Bayesian maintenance risk network, and the probability state of the Bayesian maintenance risk network is updated without setting evidence or changing any node state.

[0080] S24. By setting the event risk variable to State 1 in the GeNIe software, indicating a 100% probability of occurrence, and observing the changes in the probability of the parent variable, we can determine the most likely direct risk factor causing the accident based on its posterior probability. Once this setting is complete, we update the Bayesian maintenance risk network to obtain the posterior probability of the parent variable of the risk variable.

[0081] S26 and GeNIe software offer sensitivity analysis capabilities for Bayesian maintenance risk networks. By setting a target node, they can calculate the nodes in the BN network that most significantly influence the target node. By setting risk variables as target nodes in the BN model for sensitivity analysis, we can understand the impact of their causes on accident occurrence. Sensitivity analysis experiments are conducted using key events derived from forward reasoning as target nodes.

[0082] S3. Based on the results of risk analysis, a dynamic evolution model is constructed, and decision analysis is performed based on the dynamic evolution model.

[0083] Based on the results of the risk analysis, maintenance error factors were categorized into technical and non-technical errors. For technical errors, corrections were made based on the maintenance manual. For non-technical errors, a dynamic evolution model was constructed, covering the entire maintenance organization process, from maintenance plan development, execution, inspection, to operation and supervision, involving maintenance personnel, equipment, management, and supervision. Decision analysis was then conducted based on the dynamic evolution model.

[0084] A further implementation is that the function of repairing the fault is expressed as:

[0085] F(t)=F(t-dt)+R F (t)-fault rectification rate (t)*dt, (5)

[0086] R F (t)=k7*(1-P(t))*(1-S(t)*(1-C(t)), (6)

[0087] Where F(t) represents the number of maintenance failures; F(t-dt) represents the historical cumulative number of maintenance failures at time point t-dt; R F (t) represents the failure rate; k7 represents the influence coefficient of the combined effects of the maintenance procedure perfection P(t), personnel skill level S(t), and inspection adequacy C(t); P(t) represents the maintenance procedure perfection; S(t) represents the personnel skill level.

[0088] A further implementation is that the causal loop of the dynamic evolution model includes four feedback loops.

[0089] Inaccurate or incomplete procedural information makes it difficult for maintenance personnel to correctly execute maintenance procedures. Inadequate execution of maintenance procedures increases the probability of equipment failure, leading to an increase in the number of maintenance failures. Increased failures prompt organizations to provide experience feedback, but if the feedback mechanism is imperfect, it may not be able to effectively improve maintenance procedures. Due to insufficient feedback, maintenance procedures cannot be updated and improved in a timely manner, and procedural information still has problems such as missing, inconsistent, or incorrect information, forming a positive feedback loop. The function is expressed as:

[0090] P(t)=P(t-dt)+R p (t)*dt, (7)

[0091] R p (t)=k3*(P 目标 -P(t))+k4*A(t), (8)

[0092] Where P(t) represents the degree of perfection of the maintenance program; P(t-dt) represents the historical cumulative degree of perfection of the maintenance program at time point t-dt; R p (t) represents the program improvement rate; k3 represents the influence coefficient of the maintenance program perfection degree; k4 represents the influence coefficient of the program information accuracy; P 目标 It represents the expected degree of perfection of the maintenance procedure; A(t) represents the accuracy of the procedure information.

[0093] Lack of adequate training for maintenance personnel will lead to insufficient professional skills and knowledge. Insufficient skills will affect the quality of maintenance personnel's execution of maintenance procedures and inspection work. Failure to strictly follow maintenance procedures or insufficient inspections will lead to an increase in the number of maintenance failures. Increased failures will cause the organization to question the ability of maintenance personnel. However, without an effective training and improvement mechanism, personnel capabilities will not be improved, forming a positive feedback loop. The function is expressed as:

[0094] S(t)=S(t-dt)+R S (t)*dt, (9)

[0095] R s (t) = k1*(S 目标 -S(t))+k2*G(t), (10)

[0096] In the formula, S(t) represents the skill level of personnel; S(t-dt) represents the historical cumulative skill level of personnel at the time point t-dt; R S (t) represents the training rate; k1 represents the influence coefficient of personnel skill level; k2 represents the influence coefficient of communication effect; S 目标 represents the expected skill level of personnel; G(t) represents the communication effect.

[0097] Without adequate supervision, it is impossible to ensure that maintenance procedures are strictly implemented, and inspection and acceptance work will become a formality. Ineffective maintenance procedures and lax inspection and acceptance will increase the number of maintenance failures. Increased failures will put greater pressure on the supervisor. However, if there are problems with the supervision system itself, the supervision effect will still be poor, and it will be impossible to effectively improve the implementation of maintenance procedures and inspection and acceptance, forming a positive feedback loop. The function is expressed as:

[0098] M(t)=M(t-dt)+R M (t)*dt, (11)

[0099] R M (t)=k5*(M 目标 -M(t))+k6*F(t), (12)

[0100] Where M(t) represents the supervision intensity; M(t-dt) represents the historical cumulative supervision intensity at time point t-dt; M 目标 Indicates the expected supervision intensity; R M (t) represents the supervision strengthening rate; k5 represents the supervision intensity influence coefficient, and k6 represents the maintenance failure number influence coefficient.

[0101] Auxiliary variables: process information accuracy A(t), inspection adequacy C(t), and communication effectiveness G(t) are as follows:

[0102] A(t)=P(t)*k8, (13)

[0103] C(t)=C(t)*k9, (14)

[0104] G(t)=S(t)*k 10 , (15)

[0105] Where, k8 represents the influence coefficient of the degree of perfection of maintenance procedures on the accuracy of procedure information; k9 represents the influence coefficient of supervision intensity on the adequacy of inspection; k 10 It represents the influence coefficient of factors such as personnel training intensity and organizational management on the accuracy of communication effectiveness.

[0106] Among them, F(t), P(t), S(t), and M(t) are state variables, and R F 、R p 、R s 、R M As rate variables, A(t), C(t), G(t) as auxiliary variables, k1-k 10 As constants, based on the causal loop and relationship mechanism of the maintenance error decision-making process, state variables, rate variables, auxiliary variables and constants are extracted, and the causal relationship is modeled using VENSIM software to establish a system flow chart for the accident emergency response process.

[0107] Example 2:

[0108] This embodiment provides a system for constructing a civil aviation maintenance risk prediction framework, including: a data acquisition module for collecting multi-source maintenance data and maintenance manual standards to obtain a multi-source data set; a model construction module for classifying events and risk variables based on the multi-source data set to obtain Bayesian network variables; and constructing a Bayesian maintenance risk network based on the Bayesian network variables, which is used for risk analysis; and a decision analysis module for constructing a dynamic evolution model based on the results of the risk analysis, and performing decision analysis based on the dynamic evolution model.

[0109] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for constructing a civil aviation maintenance risk prediction framework, characterized in that: The following steps are involved: S1, collect multi-source maintenance data and maintenance manual standards to obtain a multi-source data set; S2. Classifying events and risk variables based on the multi-source data set to obtain Bayesian network variables; and constructing a Bayesian maintenance risk network based on the Bayesian network variables, wherein the Bayesian maintenance risk network is used for risk analysis; S3. Based on the results of the risk analysis, a dynamic evolution model is constructed, and a decision analysis is performed based on the dynamic evolution model.

2. A method for constructing a civil aviation maintenance risk prediction framework according to claim 1, characterized in that: The Bayesian network variables include: events, environmental factors, installation and commissioning errors, maintenance procedures and management deficiencies, parts and tool management omissions, personnel and training deficiencies, documentation and compliance issues, supervision and resource issues, and time and technology pressures.

3. The method for constructing a civil aviation maintenance risk prediction framework according to claim 1, characterized in that: The method for constructing the Bayesian maintenance risk network based on the Bayesian network variables in S2 includes: identifying dependency relationships of the Bayesian network variables based on structure learning, and constructing the Bayesian maintenance risk network based on the dependency relationships; Including: S21, for each node of the Bayesian maintenance risk network, obtain the joint probability distribution based on parameter learning: Where: P(R1,…,R n , C) represents the joint probability distribution; the severity C of the maintenance event is the risk variable set R={R1,...,R i } is the only parent variable; P(C) represents the probability of maintenance events; n represents the number of risk variables; P(R i |C) represents the probability of risk variables under maintenance events; S23. Introduce a tree structure for risk variables, add additional edges from data learning to represent the dependencies between risks, and obtain the functional relationship: In the formula, {C,R π(i) } represents the relationship between maintenance event C and a risk variable R i The parent node R π(i) The relationship between them; {C} represents the maintenance event; π(i) represents the function π that converts each risk variable R i Mapped to its parent node; S25. Based on the joint probability distribution and the functional relationship, obtain the posterior probability of the severity of the maintenance event: Where, P(C|R1,…,R n ) represents the known risk variables R1,...,R n The probability of maintenance event C occurring under the conditions of their respective specific states; P(R root |C) represents the root node risk variable R root Conditional probability under a given class variable C state; P(R i |C,∏R i ) represents the risk variable R i Given a class variable C and its parent node set ΠR i Conditional probability under state; S27. Based on the given severity of the maintenance event, the dependency relationship between the two risk variables is obtained: Where, I P represents the conditional mutual information between two risk variables; r ii Represents the risk variable R i The i-th state; r ji Represents the risk variable R j The i-th state; c i Indicates the i-th state of the severity of the maintenance event; b is a constant; P(r ii ,r ji |c i ) means that given C=c i Under the condition of i and R j The joint conditional probability of P(r ii |c i ) means that given C=c i Under the condition of i The marginal conditional probability of P(r ji |c i ) means that given C=c i Under the condition of j The marginal conditional probability of .

4. A method for constructing a civil aviation maintenance risk prediction framework according to claim 3, characterized in that: The method for performing risk analysis based on the Bayesian maintenance risk network in S2 includes: S22. Calculate the probability of occurrence of accident risk factors for the Bayesian maintenance risk network using the software GeNIe, and update the probability state of the Bayesian maintenance risk network without providing evidence or changing any node state; S24, by setting the state of the event risk variable to State1 in the GeNIe software, observing the probability change of the parent variable, and updating the Bayesian maintenance risk network to obtain the posterior probability of the parent variable of the risk variable; S26. Setting the risk variable as the target node in the Bayesian maintenance risk network to perform sensitivity analysis.

5. The method for constructing a civil aviation maintenance risk prediction framework according to claim 1, characterized in that: Based on the results of risk analysis, the maintenance error factors are classified into: technical errors and non-technical errors; For technical errors, make corrections based on the maintenance manual; For non-technical errors, the dynamic evolution model is constructed, and decision analysis is performed based on the dynamic evolution model.

6. The method for constructing a civil aviation maintenance risk prediction framework according to claim 1, characterized in that: Maintenance faults are indicated as: F(t)=F(t-dt)+R F (t)-fault rectification rate (t)*dt, R F (t)=k7*(1-P(t))*(1-S(t)*(1-C(t)), Where F(t) represents the number of maintenance failures; F(t-dt) represents the historical cumulative number of maintenance failures at time point t-dt; R F (t) represents the failure rate; k7 represents the influence coefficient of the combined effects of the maintenance procedure perfection P(t), personnel skill level S(t), and inspection adequacy C(t); P(t) represents the maintenance procedure perfection; S(t) represents the personnel skill level.

7. A method for constructing a civil aviation maintenance risk prediction framework according to claim 6, characterized in that: The causal loop of the dynamic evolution model includes four feedback loops: P(t)=P(t-dt)+R p (t)*dt, R p (t)=k3*(P 目标 -P(t))+k4*A(t), Where P(t) represents the degree of perfection of the maintenance program; P(t-dt) represents the historical cumulative degree of perfection of the maintenance program at time point t-dt; R p (t) represents the program improvement rate; k3 represents the influence coefficient of the maintenance program perfection degree; k4 represents the influence coefficient of the program information accuracy; P 目标 It represents the expected degree of perfection of the maintenance program; A(t) represents the accuracy of the program information; S(t)=S(t-dt)+R S (t)*dt, R s (t)=k1*(S 目标 -S(t))+k2*G(t), In the formula, S(t) represents the skill level of personnel; S(t-dt) represents the historical cumulative skill level of personnel at the time point t-dt; R S (t) represents the training rate; k1 represents the influence coefficient of personnel skill level; k2 represents the influence coefficient of communication effect; S 目标 It represents the expected skill level of personnel; G(t) represents the communication effect; M(t)=M(t-dt)+R M (t)*dt, R M (t)=k5*(M 目标 -M(t))+k6*F(t), Where M(t) represents the supervision intensity; M(t-dt) represents the historical cumulative supervision intensity at time point t-dt; M 目标 Indicates the expected supervision intensity; R M (t) represents the supervision strengthening rate; k5 represents the supervision intensity influence coefficient, and k6 represents the maintenance failure number influence coefficient; And auxiliary variables: process information accuracy A(t), inspection adequacy C(t), communication effectiveness G(t): A(t)=P(t)*k8, C(t)=C(t)*k9, G(t)=S(t)*k 10 , Where, k8 represents the influence coefficient of the degree of perfection of maintenance procedures on the accuracy of procedure information; k9 represents the influence coefficient of supervision intensity on the adequacy of inspection; k 10 It represents the influence coefficient of factors such as personnel training intensity and organizational management on the accuracy of communication effectiveness.

8. A system for constructing a civil aviation aircraft maintenance risk prediction framework, the system being used to implement the method according to any one of claims 1 to 7, characterized in that: include: A data acquisition module is used to collect multi-source maintenance data and maintenance manual standards to obtain a multi-source data set; A model building module, configured to classify events and risk variables based on the multi-source data set to obtain Bayesian network variables; and constructing a Bayesian maintenance risk network based on the Bayesian network variables, wherein the Bayesian maintenance risk network is used for risk analysis; The decision analysis module is used to construct a dynamic evolution model based on the results of the risk analysis and perform decision analysis based on the dynamic evolution model.

Images