Aviation equipment risk assessment method of dynamic three-branch FMEA based on risk interaction

Through the dynamic three-branch FMEA method, combining multi-stage risk information, an incomplete information system is established, and the risk prospect utility value and attribute weight are calculated, the problem that risk assessment methods in the existing technology are difficult to combine multi-channel information, and the aviation equipment is realized Accurate analysis and dynamic management of security risks.

CN118982229APending Publication Date: 2024-11-19AIR FORCE UNIV PLA
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410976420.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-20
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The existing aviation equipment risk assessment methods mainly focus on a single stage of the equipment life cycle. Failure to effectively combine information from different channels makes it difficult to meet the accurate analysis of equipment safety risks in complex decision-making environments.

Method used

Using a risk assessment method based on dynamic three-branch FMEA, we can collect and process multiple risk information in the design stage, mechanical simulation test and service stage, establish an incomplete information system, calculate the risk prospect utility value, and consider the subjective factors of decision makers and information entropy. , determine and cluster risk attribute weights, and finally arrange the key components of the equipment risk.

Benefits of technology

It has achieved a more objective, real and accurate prediction and analysis of aviation equipment quality risks, and can dynamically respond to equipment changes and provide scientific and effective decision-making basis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982229A_ABST
    Figure CN118982229A_ABST
Patent Text Reader

Abstract

The invention discloses an aviation equipment risk assessment method of dynamic three-branch FMEA based on risk interaction. The method comprises the following steps: collecting and processing multi-aspect failure mode information; calculating weights of different condition attributes under the three granularities; calculating a risk foreground utility value; clustering the similar risk attributes; calculating a utility matrix of each component under different state decisions; and sorting the risk of each component. According to the method disclosed by the invention, the information in the prior art, in the event and after the event is used as the input of risk assessment, so that the risk probability of the current equipment quality can be reflected more intuitively and truly; meanwhile, during risk analysis, the interaction relation between risk factors, the attitude of a decision maker to the risk and the dynamic analysis of key risk components are considered, and a more scientific and effective basis is provided for management decision making.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the fields of weapon equipment, risk assessment and three-way decision-making, and more specifically, to an aviation equipment risk assessment method based on dynamic three-way FMEA. Background Art

[0002] Aviation equipment is the most solid material foundation for a country to obtain "air supremacy". As far as its development is concerned, ensuring the safety of aviation equipment in service is an extremely important matter. Since aviation equipment is essentially an integrated system composed of single structural components with different functions, its development, production, use and maintenance processes are quite complicated. At the same time, these high-end and complex aviation equipment are expensive. Once quality problems occur, it will not only bring huge economic losses, but also cause casualties of combatants, and even pose a serious threat to national security. Therefore, in order to more clearly understand the quality risks of these aviation equipment and to conduct effective risk prediction and control, it is necessary to conduct a comprehensive and systematic analysis, quantification and evaluation of the key risk components of the single structure of these aviation equipment.

[0003] At present, the research on aviation equipment risk assessment based on failure mode and effect analysis (FMEA) generally focuses on analyzing the potential risks in a single stage of the equipment life cycle, such as analyzing the failure mode in advance, and most of them only use one-sided information to improve the FMEA method in a variety of different ways. Few scholars combine information from different channels. However, with the increasing complexity of the decision-making environment, the environment in which aviation equipment is located is also constantly changing. It is difficult to meet the needs of accurately analyzing equipment safety risks under actual combat training conditions by only identifying key risks through one-sided safety data in a single life stage of the equipment. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an aviation equipment risk assessment method based on a dynamic three-branch FMEA of risk interaction, and the specific steps are as follows:

[0005] Step 1: Collect and process various failure mode information;

[0006] Through investigation, we obtain risk information from three aspects: FMEA data in the design stage, mechanical simulation test data, and failure damage data in the service stage. We pre-process the information and convert all of it into factor values ​​that have a positive impact on risk analysis. Then, we establish an incomplete information system for the risk of single structural components and measure the incomplete information system.

[0007] The specific implementation process of information preprocessing is as follows:

[0008] (1) For the FMEA data in the design phase, the "severity level" and "probability level" are used as indicators for the pre-analysis of component structure risk analysis. The language terms of "severity level" and "probability level" are converted into certain numbers between 0 and 1; according to the levels included in "severity" and "probability of occurrence", the corresponding number of certain numbers is set. The higher the level, the larger the corresponding certain value, as shown in the following table;

[0009] Severity level Determine the number of indicators Probability level Determine the number of indicators Ⅰ 0.9 A 0.9 Ⅱ 0.5 B 0.65 Ⅲ 0.3 C 0.4 Ⅳ 0.1 D 0.15

[0010] (2) For the mechanical simulation test data information, "force magnitude" and "fracture coefficient" are used as indicators for in-process component structure risk analysis. To simplify the problem, the relationship between "force magnitude" and "fracture coefficient" is determined according to the following formula;

[0011]

[0012] In the formula, Q is the fragility coefficient, F is the force magnitude, E is the elastic modulus of the material, A is the cross-sectional area of ​​the material, L is the length of the material, q1 and q2 are constant coefficients, which depend on the specific experimental conditions and material properties;

[0013] (3) Based on the information of the failure damage data in the service stage, the “accident occurrence ratio” and “severity” are used as indicators for the post-analysis of component structure risk analysis, and the language terms of “accident occurrence ratio” and “severity” are converted into a definite number between 0 and 1; among which, the accident occurrence ratio is defined as the ratio of single component failure problems to total component failure problems, and the “severity” is determined according to the degree of harm of the failure problem. The more serious the degree of harm, the larger the corresponding definite value, as shown in the following table;

[0014] part Failure issues Accident rate Severity Indicator A crack 0.5 0.7 B Broken screw 0.3 0.6 C corrosion 0.2 0.5 D Wear 0.1 0.4

[0015] (4) Establish an incomplete risk information system for monomer structural components. The specific implementation process is as follows:

[0016] Let the decision object x i The set is: U = {x1, x2, ... x m}, m is the number of decision objects, and the attribute feature set of the evaluation decision object is: A = {a1, a2, ... a n}, n is the number of attribute features, and at the same time, decisions are made on information of k granularities, corresponding to k decision information systems, thus forming an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), where U k is a non-empty set of decision objects, C kis a non-empty finite set of conditional attributes, g k is a non-empty finite set of decision attributes, C k ∪g k Represents the set C k With the collection g k The union of k is a set of attribute values, f k For each object attribute value, through U k and C k ∪g k The Cartesian product of k The mapping function is obtained: f k :U k ×(C k ∪g k ) → A k ;

[0017] The specific implementation process of measuring incomplete information systems is as follows:

[0018] In an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), if the attribute value is a Boolean value, then the i-th decision object x at the k-th information granularity i The attribute value of the jth attribute Its vision Distance As shown below:

[0019]

[0020] If the attribute value is a certain number, the distance between the attribute value and the vision value As shown below:

[0021]

[0022] If the attribute is an interval number, the distance between the attribute value and the vision value As shown below:

[0023]

[0024] In the formula, is the attribute value interval, Respectively represent the lower and upper bounds of the attribute value interval;

[0025] Step 2: Calculate the risk prospect utility value;

[0026] According to the distance between the attribute value and the risk vision value, the risk loss matrix is ​​calculated. Then, considering the subjective factors of personnel in the risk analysis process, the risk loss matrix is ​​optimized using the prospect value function and converted into the actual expected utility value. That is, the larger the expected utility value, the smaller the corresponding risk value.

[0027] The specific implementation process of the risk loss matrix is ​​as follows:

[0028] According to the distance between the attribute value and the risk prospect value, the risk loss is defined, and the loss function corresponding to the three decision models under each attribute is established. Then the reference point is determined to obtain the profit and loss matrix:

[0029]

[0030] The specific implementation process of converting to actual expected utility value is as follows:

[0031] Using the prospect value function V(x) in prospect theory, the profit and loss matrix is ​​mapped using the prospect value function to obtain the true risk expected utility value.

[0032]

[0033] Among them, V(x) represents the utility or value of the prospect in the value function, x represents the gain or loss of the prospect, α and β represent the gain and loss functions respectively, and are placed in the upper right corner to represent the power series, 0≤α,β≤1, θ represents the risk aversion coefficient, and when θ>1, it means that the decision maker is more sensitive to losses than gains;

[0034] The utility value considering the risk preference of the decision maker is calculated. Taking into account the non-independence of attributes, the following λ fuzzy measure is used to fuse different attribute values:

[0035] μ(A∪B)=μ(A)+μ(B)+λμ(A)μ(B)

[0036]

[0037] In the formula, μ(·) represents the attribute value, A and B represent two different attributes, A∪B is the union of attribute A and attribute B, λ is a measure between 0 and 1, and x i represents the i-th decision object, g(x i ) is x i The corresponding λ fuzzy measure;

[0038] Step 3: Calculate the weights of the three conditional attributes of different granularities;

[0039] The three data sources are called three-granularity data, where the design granularity is FMEA data, the test granularity is mechanical test data, and the reality granularity is accident and failure data. The attribute weight at each granularity is determined by calculating the information entropy method.

[0040] Determine the attribute weight at each granularity. The specific implementation process is as follows:

[0041] In terms of the degree of influence on the attribute weight, it is actually hoped that the more incomplete information there is, the smaller the attribute weight will be. Combined with the vision value, it will be converted into an actual utility value. The present invention will use the value matrix to solve the attribute weight. For the value matrix corresponding to the incomplete information system, if the value matrix of this attribute has a larger degree of discreteness, the attribute weight will be larger, otherwise, the attribute weight will be smaller. Therefore, the definition of information entropy is introduced, and the size of information entropy is used as a measure of attribute weight.

[0042] The value entropy of the decision attribute is defined as:

[0043]

[0044] According to the entropy value of each attribute, the definition of attribute weight is given:

[0045]

[0046] In the formula, the attribute weight and K represents the maximum number of granularities or decision information systems, and J represents the number of attributes of the decision object;

[0047] Step 4: Cluster similar risk attributes;

[0048] Clustering of risks with similar attributes is done through risk utility deviation calculation, risk similarity calculation, and risk similarity classification, to simplify and accurately control risks.

[0049] The specific implementation process of risk utility deviation calculation is as follows:

[0050] First, the risk-related utility deviation of each component is calculated; the ratio of each risk actual utility value in each granular attribute to the maximum value in the attribute is defined as the risk-related utility deviation; in an incomplete mixed information system (U k ,C k ∪g k ,A k ,f k ), let Represents attribute value With vision value The relevant dispersion, the dispersion of the risk-related utility of each component Solve it by the following formula:

[0051]

[0052] In the formula, is the maximum value of the risk expected utility value among the j attributes of the i-th decision object at the k-th granularity;

[0053] Based on the obtained risk-related utility deviations of each component and the weight values ​​of each attribute at each granularity, the risk utility deviations at different granularities are obtained. As shown in the following formula;

[0054]

[0055] The specific implementation process of risk similarity calculation is as follows:

[0056] Based on the risk-related utility deviation, in an incomplete mixed information system (U k ,C k ∪g k ,A k ,f k ), let Csr k (x i ,x l ) represents the risk similarity between different components. Based on this, for any two decision objects x i ,x l ∈U, the risk similarity Csr between different decision objects at three granularities is obtained by the following cosine similarity calculation formula k (x i ,x l );

[0057]

[0058] The specific implementation process of risk similarity classification is as follows:

[0059] Decision object x i The δ-neighborhood of the same class is defined as:

[0060]

[0061] In the formula, δ is a similarity judgment value, 0≤δ≤1. When the similarity between two decision objects is greater than or equal to δ, it means that the decision object x i 、x l The risk level is the same;

[0062] On the basis of similarity classification, the proportion of the number of components with the same risk level at each granularity to the total number of components is calculated according to the following probability measurement formula ρ(x i );

[0063]

[0064] In the formula, ||U|| is the total number of decision objects, For the decision object x i The number of decision objects with the same risk level;

[0065] Step 5: Calculate the utility matrix of each component under different state decisions; abstract the determination of key risk components of the structure into a decision problem, that is, for each analyzed monomer structural component, there are three decision classifications: one is that the component is considered to be a key risk component Second, it is considered that this component is not a critical risk component Third, it is considered that the component is at the boundary between the key risk component and the non-key risk component. Similarly, for the three decision categories, each component has three actual states, that is, the component is actually a key risk component S P , not actually a key risk component S N and uncertain S B ; Based on the acquired data, the decision utility matrix of each component in different states is calculated;

[0066] Calculate the decision utility matrix of each component in different states. The specific implementation process is as follows:

[0067] First, the probability calculation of decision division is to solve the correlation between each part and the three part decision classifications through the existing k granularity data;

[0068] in (U k ,C k ∪g k ,A k ,f k ) on the represents the set of decision states, Represents the decision attribute g k The value range of Ω k The three states are defined as:

[0069]

[0070] In the formula, Indicates key risk components, A component that represents the boundary state between critical risks and non-critical risks. represents non-critical risk components; a, h, and r represent the attribute values ​​corresponding to three different decisions respectively; for three different decision object attribute states, there are three targeted decisions corresponding to them. When the decision object attribute is Accept the decision when the decision object attribute is When the decision object attribute is When decision making is made, the decision state set B = {b A ,b H ,b R}, where b A Accept the decision, b H represents delayed decision, b R Rejection of decision;

[0071] In the actual risk analysis work, the actual accident problem data or new test accidents should be used as the updated data to calculate the decision-making division probability. Since the safety risk analysis here does not involve updated data, the risk score not used in the design granularity FMEA is used as the updated data; according to the latest risk score data, different components are further divided into three levels of high risk, medium risk and low risk according to the predetermined score, so that x i Belong to the status The decision partition probability The calculation method is defined as:

[0072]

[0073] In the formula, ◇∈{A,H,R}, it means that the variable ◇ belongs to any variable among the variables A,H,R. k ,C k ∪g k ,A k ,f k ), ρ(x i ) represents the ratio of the number of components with the same risk level to the number of all components at each granularity. The ratio of the total number of components in the high / medium / low risk level to which the component belongs to the total number of all components;

[0074] Secondly, the associated utility matrix of each component in different states is calculated as:

[0075]

[0076] This matrix is ​​obtained by the following six associated loss functions;

[0077]

[0078] In the formula, Indicates that the component is actually a critical risk part, but the decision-maker considers that the component is at the boundary between the critical risk component and the non-critical risk component, resulting in losses; Indicates the loss caused by the fact that it is actually a key risk part, but the decision is that the part is not a key risk part; Indicates that the component is actually at the boundary state between the key risk component and the non-key risk component, but the decision-maker considers the component to be a key risk component and generates losses; The component is actually at the boundary between the critical risk component and the non-critical risk component, but the decision-maker considers the loss caused by the component as a non-critical risk component; Indicates that the component is actually a non-critical risk component, but the decision-maker considers the component to be a critical risk component; The component is actually a non-critical risk component, but the decision-maker believes that the component is at the boundary between the critical risk component and the non-critical risk component, resulting in losses; γ and μ are both loss magnification coefficients, is the risk utility dispersion at different granularities;

[0079] Finally, combining the decision-making probability of each component in different states and the associated utility matrix of each component in different states, the utility matrix under different part decisions is calculated according to the following formula:

[0080]

[0081] In the formula, Indicates the utility value of accepting the decision when the decision-maker considers the component as a key risk component. It represents the utility value of the delayed decision when the decision-maker considers that the component is at the boundary state between the critical risk component and the non-critical risk component. It represents the utility value of rejecting the decision when the decision considers that the component is a non-critical risk component. Indicates the probability of considering this component as a key risk component, Indicates the probability of the component being considered as a boundary state between a critical risk component and a non-critical risk component, Indicates the probability of considering the component as a non-critical risk component.

[0082] Step 6: Sort the risk of each component; divide the key parts under each granularity according to the set three-branch decision-making division rule; then use the given three granularities to assign weights, merge them, calculate the score of each key part, and finally get the risk of each component for ranking;

[0083] The specific rules for dividing the three decisions are as follows:

[0084] If satisfied: and have

[0085] If satisfied: and have

[0086] If satisfied: and have In the formula, A collection of key risk areas. is the boundary state set between key risk areas and non-key risk areas, For non-critical risk areas

[0087] Calculate the score of each key part. The specific implementation process is as follows:

[0088] The weights of multiple granularities in risk analysis are set as W = {W1, W2…W k}, where W1, W2…W k Respectively represent the weight coefficients of the 1st, 2nd…kth granularity;

[0089] Set a signal function

[0090]

[0091] In the formula, Represents the decision object x at the kth granularity i The scores under different risk positions;

[0092] Through the signal function, combined with the weights at different granularities, the final score R(x i ):

[0093]

[0094] The method of the present invention is used to confirm key risk components of equipment and to quantify their risk levels.

[0095] The advantages and effects of the present invention are as follows: compared with previous risk assessment methods, the present invention takes into account the mutual interaction between risk factors, the risk attitude of decision makers, and risk information before, during and after the event, and can predict and analyze the current equipment quality risks more objectively, truly and accurately. At the same time, with the changes of time, the key parts of aviation equipment will continue to change during the occurrence of new events. The model of the present invention can respond to such changes through continuous iteration, reflecting the dynamic nature of decision-making and providing a more scientific and effective basis for management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 The figure is a schematic diagram of the dynamic three-branch FMEA risk assessment process based on risk interaction. DETAILED DESCRIPTION

[0097] The present invention takes into account the interaction of different risk factors, the preference of different decision makers for risk attitudes, and the timing of decision making. First, the cumulative prospect theory is used to improve the loss function, and then the three-branch decision theory is introduced. For different component states (in the state of critical risk components, in the boundary state between critical risk components and non-critical risk components, and in the state of non-critical risk components) and their three possible decisions (accept decision, delay decision, and reject decision), by integrating the FMEA analysis results of the aviation equipment design and manufacturing stage (before), mechanical simulation experiments under actual combat training conditions (during), and safety issues during service (after), the decision loss function matrix of different components in different states is calculated, and finally, according to the set decision rules, the key risk components of the structure and the size of the risk are judged. A universal and highly operable safety risk assessment method is provided for aviation equipment.

[0098] According to the above principle requirements of the present invention, the specific steps are as follows:

[0099] Step 1: Collect and process various failure mode information.

[0100] Through investigation, we obtain risk information from three aspects: FMEA data (before the event), mechanical simulation test data (during the event), and failure damage data (after the event) in the service stage. We preprocess the information and convert it into factor values ​​that have a positive impact on risk analysis. Then we establish an incomplete information system for the risk of single structural components and measure the incomplete information system.

[0101] The specific implementation process of preprocessing the information in step 1 is as follows:

[0102] (1) For the FMEA data in the design phase, the present invention uses "severity level" and "probability level" as indicators for pre-analysis of component structure risk analysis, and converts the language terms of "severity level" and "probability level" into a definite number between 0 and 1. According to the level included in "severity" and "probability of occurrence", the corresponding number of definite numbers is set. The higher the level, the larger the corresponding definite value, as shown in the following table.

[0103] Severity level Determine the number of indicators Probability level Determine the number of indicators Ⅰ 0.9 A 0.9 Ⅱ 0.5 B 0.65 Ⅲ 0.3 C 0.4 Ⅳ 0.1 D 0.15

[0104] (2) For the in-process mechanical simulation test data information, the present invention adopts "force magnitude" and "fracture coefficient" as indicators for in-process component structure risk analysis. In order to simplify the problem, the relationship between "force magnitude" and "fracture coefficient" is determined according to the following formula.

[0105]

[0106] In the formula, Q is the fragility coefficient, F is the force magnitude, E is the elastic modulus of the material, A is the cross-sectional area of ​​the material, L is the length of the material, and q1 and q2 are constant coefficients that depend on the specific experimental conditions and material properties.

[0107] (3) With respect to the information of the fault damage data in the post-service stage, the present invention adopts "accident occurrence ratio" and "severity" as indicators for post-analysis of component structure risk analysis, and converts the language terms of "accident occurrence ratio" and "severity" into a definite number between 0 and 1. Among them, the accident occurrence ratio is defined as the ratio of single component failure problems to total component failure problems, and the "severity" is determined according to the degree of harm of the fault problem. The more serious the degree of harm, the larger the corresponding definite value, as shown in the following table.

[0108] part Failure issues Accident rate Severity Indicator A crack 0.5 0.7 B Broken screw 0.3 0.6 C corrosion 0.2 0.5 D Wear 0.1 0.4

[0109] (4) Establish an incomplete risk information system for monomer structural components. The specific implementation process is as follows:

[0110] Let the decision object x i The set is: U = {x1, x2, ... x m}, m is the number of decision objects, and the attribute feature set of the evaluation decision object is: A = {a1, a2, ... a n}, n is the number of attribute features, and at the same time, decisions are made on information of k granularities, corresponding to k decision information systems, thus forming an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), where U k is a non-empty set of decision objects, C k is a non-empty finite set of conditional attributes, g k is a non-empty finite set of decision attributes, C k ∪g k Represents the set C k With the collection g k The union of k is a set of attribute values, f k For each object attribute value, through U k and C k ∪g k The Cartesian product of k The mapping function is obtained: f k :U k ×(C k ∪g k ) → A k .

[0111] The specific implementation process of measuring incomplete information systems is as follows:

[0112] In an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), if the attribute value is a Boolean value, then the i-th decision object x at the k-th information granularity i The attribute value of the jth attribute Its vision Distance As shown below:

[0113]

[0114] If the attribute value is a certain number, the distance between the attribute value and the vision value As shown below:

[0115]

[0116] If the attribute is an interval number, the distance between the attribute value and the vision value As shown below:

[0117]

[0118] In the formula, is the attribute value interval, They represent the lower and upper bounds of the attribute value interval respectively.

[0119] Step 2: Calculate the risk prospect utility value.

[0120] According to the distance between the attribute value and the risk vision value, the risk loss matrix is ​​calculated. Then, considering the influence of subjective factors of personnel in the risk analysis process, the risk loss matrix is ​​optimized using the prospect value function and converted into the actual expected utility value. That is, the larger the expected utility value, the smaller the corresponding risk value.

[0121] The specific implementation process of the risk loss matrix is ​​as follows:

[0122] According to the distance between the attribute value and the risk prospect value, the risk loss is defined, and the loss function corresponding to the three decision models under each attribute is established. Then the reference point is determined to obtain the profit and loss matrix:

[0123]

[0124] The specific implementation process of converting to actual expected utility value is as follows:

[0125] Using the prospect value function V(x) in prospect theory (this function is well known to those skilled in the art), the profit and loss matrix is ​​mapped using the prospect value function to obtain the true risk expected utility value.

[0126]

[0127] Among them, V(x) represents the utility or value of the prospect (i.e., possible future results) in the value function, x represents the gain or loss of the prospect, α and β represent the gain and loss functions respectively, and are placed in the upper right corner to represent the power series, 0≤α,β≤1, θ represents the risk aversion coefficient, and when θ>1, it means that the decision maker is more sensitive to losses than to gains.

[0128] The utility value considering the risk preference of the decision maker is calculated. Taking into account the non-independence of attributes, the following λ fuzzy measure is used to fuse different attribute values:

[0129] μ(A∪B)=μ(A)+μ(B)+λμ(A)μ(B)

[0130]

[0131] In the formula, μ(·) represents the attribute value, A and B represent two different attributes, λ is a measure between 0 and 1, and x i represents the i-th decision object, g(x i ) is x i The corresponding λ fuzzy measure.

[0132] Step 3: Calculate the weights of the three conditional attributes with different granularities.

[0133] The three data sources are called three-granularity data, among which the design granularity is FMEA data, the test granularity is mechanical test data, and the actual granularity is accident and failure data. The attribute weight at each granularity is determined by calculating the information entropy.

[0134] Determine the attribute weight at each granularity. The specific implementation process is as follows:

[0135] In terms of the degree of influence on attribute weights, it is actually hoped that the more incomplete information there is, the smaller the attribute weight will be. Combined with the vision value, it will be converted into an actual utility value. The present invention will use the value matrix to solve the attribute weights. For the value matrix corresponding to the incomplete information system, if the value matrix of this attribute has a greater degree of discreteness, the attribute weight will be greater. Conversely, the attribute weight will be smaller. Therefore, the definition of information entropy is introduced, and the size of information entropy is used as a measure of attribute weight.

[0136] The value entropy of the decision attribute is defined as:

[0137]

[0138] According to the entropy value of each attribute, the definition of attribute weight is given:

[0139]

[0140] In the formula, the attribute weight and K represents the maximum number of granularities or decision information systems, and J represents the number of attributes of the decision object.

[0141] Step 4: Cluster similar risk attributes.

[0142] By calculating risk utility deviation, risk similarity and classifying risks into similar categories, similar risks with different attributes are clustered to achieve risk simplification and precise control.

[0143] The specific implementation process of risk utility deviation calculation is as follows:

[0144] First, the risk-related utility deviation of each component is calculated. The ratio of the actual utility value of each risk in each granular attribute to the maximum value of the attribute is defined as the risk-related utility deviation. k ,C k ∪g k ,A k ,f k ), let Represents attribute value With vision value The relative dispersion, therefore, the relative dispersion of the risk-related utilities of each component Solve it by the following formula:

[0145]

[0146] In the formula, It is the maximum value of the risk expected utility value among the j attributes of the ith decision object at the kth granularity.

[0147] According to the obtained risk-related utility deviation of each component and the weight value of each attribute at each granularity, the risk utility deviation at different granularities can be obtained. As shown in the following formula.

[0148]

[0149] The specific implementation process of risk similarity calculation is as follows:

[0150] Based on the risk-related utility deviation, in an incomplete mixed information system (U k ,Ck ∪g k ,A k ,f k ), let Csr k (x i ,x l ) represents the risk similarity between different components. Based on this, for any two decision objects x i ,x l ∈U, the risk similarity Csr between different decision objects at three granularities is obtained by the following cosine similarity calculation formula k (x i ,x l ).

[0151]

[0152] The specific implementation process of risk similarity classification is as follows:

[0153] Decision object x i The δ-neighborhood of the same class is defined as:

[0154]

[0155] In the formula, δ is a similarity judgment value, 0≤δ≤1. When the similarity between two decision objects is greater than or equal to δ, it means that the decision object x i 、x l The risk level is the same.

[0156] On the basis of similarity classification, the proportion of the number of components with the same risk level at each granularity to the total number of components is calculated according to the following probability measurement formula ρ(x i ).

[0157]

[0158] In the formula, ||U|| is the total number of decision objects, For the decision object x i The number of decision objects with the same risk level.

[0159] Step 5: Calculate the utility matrix of each component under different state decisions. The present invention abstracts the determination of key risk components of the structure into a decision problem, that is, for each analyzed single structural component, there are three decision classifications: one is that the component is considered to be a key risk component Second, it is considered that this component is not a critical risk component Third, it is considered that the component is at the boundary between the key risk component and the non-key risk component. Similarly, for the three decision categories, each component has three actual states, that is, the component is actually a key risk component SP , Actually not a key risk component S N and uncertain S B Based on the acquired data, the decision utility matrix of each component in different states is calculated.

[0160] Calculate the decision utility matrix of each component in different states. The specific implementation process is as follows:

[0161] First, the decision division probability calculation is to use the existing k granular data to solve the correlation between each part and the three part decision classifications.

[0162] in (U k ,C k ∪g k ,A k ,f k ) on the represents the set of decision states, Represents the decision attribute g k The value range of Ω k The three states are defined as:

[0163]

[0164] In the formula, Indicates the key risk components, A component that represents the boundary state between critical risks and non-critical risks. Indicates non-critical risk components. a, h, and r represent the attribute values ​​corresponding to three different decisions. For three different decision object attribute states, there are three targeted decisions corresponding to them. When the decision object attribute is Accept the decision when the decision object attribute is When the decision object attribute is When decision is made, we can get the decision state set B = {b A ,b H ,b R}, where b A Accept the decision, b H represents delayed decision, b R Rejection of decision.

[0165] In actual risk analysis, the acquired real accident problem data or new test accidents should be used as updated data to calculate the decision-making probability. Since the safety risk analysis here does not involve updated data, the risk score not used in the design granularity FMEA is used as updated data. According to the latest risk score data, different components are further divided into three levels of high risk, medium risk and low risk according to the predetermined score. i Belong to the status The decision partition probability P N (D k |x i ) is calculated as:

[0166]

[0167] In the formula, ◇∈{A,H,R}, it means that the variable ◇ belongs to any variable among the variables A,H,R. k ,C k ∪g k ,A k ,f k ), ρ(x i ) represents the ratio of the number of components with the same risk level to the number of all components at each granularity. It is the ratio of the total number of components in the high / medium / low risk level to the total number of all components.

[0168] Secondly, the associated utility matrix of each component in different states is calculated as:

[0169]

[0170] This matrix is ​​obtained by the following six associated loss functions.

[0171]

[0172] In the formula, It means that the component is actually a key risk part, but the decision-maker considers that the component is at the boundary state between the key risk component and the non-key risk component, resulting in losses. It indicates that the loss is caused by the fact that it is actually a key risk part, but the decision is made that the part is not a key risk part. It means that the component is actually in the boundary state between the key risk component and the non-key risk component, but the decision considers that the component is the loss caused by the key risk component. This component is actually the boundary state between the critical risk component and the non-critical risk component, but the decision-maker considers that the loss caused by this component is a non-critical risk component. It means that the component is actually a non-critical risk component, but the decision maker considers it to be a critical risk component and causes losses. The component is actually a non-critical risk component, but the decision-maker believes that the component is at the boundary between the critical risk component and the non-critical risk component, resulting in losses. γ and μ are both loss magnification factors. is the risk utility deviation at different granularities.

[0173] Finally, combining the decision-making probability of each component in different states and the associated utility matrix of each component in different states, the utility matrix under different part decisions is calculated according to the following formula:

[0174]

[0175] In the formula, Indicates the utility value of accepting the decision when the decision-maker considers the component as a key risk component. It represents the utility value of the delayed decision when the decision-maker considers that the component is at the boundary state between the critical risk component and the non-critical risk component. It represents the utility value of rejecting the decision when the decision considers that the component is a non-critical risk component. Indicates the probability of considering this component as a key risk component, Indicates the probability of the component being considered as a boundary state between a critical risk component and a non-critical risk component, Indicates the probability of considering the component as a non-critical risk component.

[0176] Step 6: Sort the risk of each component. According to the set three-branch decision-making division rules, the key parts under each granularity are divided. Then the given three granularities are given weights, which are merged to calculate the score of each key part, and finally the risk of each component is ranked.

[0177] The specific rules for dividing the three decisions are as follows:

[0178] If satisfied: and have

[0179] If satisfied: and have

[0180] If satisfied: and have In the formula, A collection of key risk areas. is the boundary state set between key risk areas and non-key risk areas, For non-critical risk areas

[0181] Calculate the score of each key part. The specific implementation process is as follows:

[0182] The weights of multiple granularities in risk analysis are set as W = {W1, W2…W k}, where W1, W2…W k They represent the weight coefficients of the 1st, 2nd…kth granularities respectively.

[0183] Set a signal function

[0184]

[0185] In the formula, Represents the decision object x at the kth granularity i The scores under different risk status.

[0186] Through the signal function and the weights at different granularities, the final score R(x i ):

[0187] Specific embodiments

[0189] The present invention is a dynamic three-branch FMEA risk assessment method based on risk interaction. Figure 1 The following is a schematic diagram of the evaluation process to further explain the specific implementation steps:

[0190] Step 1: Collect and process various failure mode information.

[0191] First, each part of the data is preprocessed, and all the attributes are converted into factor values ​​that have a positive impact on risk analysis. In this embodiment, in order to facilitate the display of research results, it is assumed that 7 key risk locations on aviation equipment are analyzed.

[0192] Q1: FMEA data preprocessing in the design phase. The severity and probability levels of multiple failure modes of a single component are averaged to form the overall severity level and probability level of failure of the component. The FMEA data obtained after processing is shown in the following table:

[0193] part Severity level Severity Rating Index Probability level A Ⅲ 0.7 0.15 B Ⅱ 0.5 0.15 C Ⅲ 0.7 0.15 D Ⅰ 0.9 0.4 E Ⅰ 0.9 0.65 F Ⅳ 0.1 0.15 G Ⅳ 0.1 0.65

[0194] Q2: Preprocessing of accident failure data during the post-service phase. The data after processing is shown in the following table:

[0195] part Failure issues Problem occurrence ratio Severity Indicator A crack 0.43 0.7 C Broken screw 0.24 0.6 E corrosion 0.12 0.5 G Wear 0.21 0.4

[0196] Q3: Integrate the processed data from the three parts and establish a single-structure risk incomplete information system based on it as shown in the following table:

[0197]

[0198] Step 2: Calculation of risk loss matrix. For the three granularities, the expectation for these attributes is that the smaller the risk, the better. Taking the severity level at the design granularity as an example, its expected risk rating is level IV, and the corresponding determination number is 0.1, that is, the expectation of this attribute is 0.1; and considering that some data under each attribute is missing, the expected value is used to fill the missing data, in order to minimize the impact of incomplete information on the subsequent risk analysis results; for the actual granularity, firstly, because this part of the data is statistical data, it is different from the actual data of the design granularity and the test granularity, and secondly, if there is no problem with the single structural component during the actual service process, the probability of the problem is 0, and the corresponding problem severity is also 0, so the data in the actual granularity is not adjusted.

[0199] According to the distance formula between the attribute value and the vision, the risk loss under the attribute value of each structural part is as follows:

[0200]

[0201] Calculation of actual expected utility value: Based on prospect theory, taking into account the influence of subjective factors of personnel in the risk analysis process, that is, analysts are more sensitive to risks, the risk-return-loss matrix is ​​optimized using the prospect value function, and the actual expected utility value is calculated. In this embodiment, α=0.32, β=0.38, θ=0.32 are set according to experience, and the risk prospect utility matrix of each component is obtained as follows: (Parameters can be adjusted according to the risk curve)

[0202]

[0203]

[0204] Step 3: Use the method of calculating information entropy to solve the weight of each granularity. The larger the information entropy of a granularity, the greater the degree of information confusion at that granularity. For risk analysis, it means that the granularity contains more risk information, so the weight is greater. After calculation, the weights of each attribute of the three granularities are:

[0205] Attribute weights for design granularity: [0.56, 0.44]

[0206] Attribute weights for test granularity: [0.39, 0.61]

[0207] Attribute weights of reality granularity: [0.53, 0.47]

[0208] Step 4: Cluster similar risk attributes. Cluster similar risks with different attributes by means of risk utility deviation calculation, risk similarity calculation, and risk similarity classification to achieve risk simplification and precise control.

[0209] Q1: Calculation of risk utility deviation. After obtaining the risk prospect utility of each component, the ratio of each value in each granular attribute to the maximum value of the attribute is calculated according to the following formula to obtain the risk-related utility deviation of each component as shown in the following subscript.

[0210]

[0211]

[0212]

[0213] Q2: Risk similarity calculation. Based on the risk-related utility deviation, the risk similarity between different structural components at three granularities is calculated according to the following formula.

[0214]

[0215] Q3: Classification of risk similarity. According to the following formula, δ=0.9 is set, that is, when the risk similarity of any two structural parts is greater than 0.9, the two parts are considered to be in the same risk similarity class. The classification results are as follows:

[0216]

[0217]

[0218] On the basis of similarity classification, the proportion of similarity classes of each structural component to the total number of components at each granularity is calculated according to the following formula, and the results are as follows:

[0219]

[0220]

[0221]

[0222] Step 5: Calculate the utility matrix of each component under different state decisions.

[0223] Q1: Calculate the probability of decision division for each component in different states. In actual risk analysis work, the actual accident problem data or new test accidents should be used as updated data to calculate the correlation. Since the safety risk analysis here does not involve updated data, in this embodiment, the risk score not used in the design granularity FMEA is used as updated data. For this score, it is also necessary to convert it into an attribute value that is positively correlated with the risk. The results are shown in the following table:

[0224] Location Risk Scoring Location Risk Scoring A 4 E 11 B 2.5 F 0 C 4 G 0 D 9

[0225] Secondly, the risk score is further divided into three levels: high risk, medium risk and low risk. The score above 8 is set as high risk, 2-8 is set as medium risk, and below 2 is set as low risk. The proportions of the three levels of risk to the total are: [0.285, 0.430, 0.285], and the probability of key part decision division at different granularities is calculated by combining the following formula.

[0226]

[0227] Take the calculation of the probability of three decision divisions at the design granularity of component "A" as an example.

[0228] The results obtained according to the above formula are as follows:

[0229]

[0230] Q2: Calculate the associated utility matrix of each component in different states. Based on the obtained risk-related utility deviation table of each component, combined with the weight value of each attribute at each granularity, the risk utility deviation at a single granularity is obtained as shown in the table:

[0231]

[0232] Based on the risk utility deviation of a single granularity, the associated utility matrix is ​​calculated according to six associated loss functions, where parameters μ = 0.4 and γ = 0.5. The utility expectation matrix under three key part decisions at three granularities is obtained. Taking part A at the design granularity as an example, the calculation process is as follows:

[0233] I BP =0.4*(1-0.5)=0.2

[0234] I PB =0.5*0.5=0.25

[0235] I NP =1-0.5=0.5

[0236] I NB =0.5*(1-0.5)=0.25

[0237] I PN =0.5

[0238] I BN =0.4*0.5=0.5

[0239] Therefore, the associated utility matrix of component A at three granularities is shown in the following table:

[0240] Different particle sizes <![CDATA[I BP ]]> <![CDATA[I PB ]]> <![CDATA[I NP ]]> <![CDATA[I NB ]]> <![CDATA[I PN ]]> <![CDATA[I BN ]]> Design granularity 0.20 0.25 0.50 0.25 0.50 0.20 Test particle size 0.40 0 1 0.50 0 0 Reality Granularity 0 0.50 0 0 1 0.40

[0241] Q3: Calculate the utility matrix of each component under different state decisions.

[0242] Taking A under the design granularity as an example, the calculation process is as follows:

[0243] L P =0.25*0.50+0.50*0.60=0.435

[0244] L B =0.20*0.60+0.20*0.60=0.270

[0245] L N =0.50*0.60+0.25*0.50=0.490

[0246] Therefore, the utility matrix under the three key position decisions is shown in the following table:

[0247]

[0248] Rank the importance or risk of key components.

[0249] Step 6: After obtaining the utilities of three different decisions for key parts of the structure at three granularities, the key parts at each granularity are divided as follows according to the set decision rules:

[0250]

[0251] After obtaining the division of key parts at each granularity, the weighted average of the three granularities is summed. Considering that the FMEA data and structural mechanics test data in the design stage are relatively complete, but there are fewer accidents and failures in the actual service stage, the design granularity weight is set to 0.3, the test granularity weight is set to 0.5, and the actual granularity weight is set to 0.2. The final score of the key parts of the structure is as follows:

[0252] Structural part name Design granularity Test particle size Reality Granularity Risk Scoring A 0 -1 -1 -0.70 B 0 0 -1 -0.20 C 0 -1 -1 -0.70 D 0 -1 -1 -0.70 E 1 -1 -1 -0.40 F -1 -1 1 -0.60 G -1 -1 0 -0.80

[0253] In the table, the higher the risk score of a structural part, the higher the risk of that part. Finally, the risk ranking of key parts is: B, E, F, A / C / D, G.

Claims

1. A dynamic three-branch FMEA risk assessment method for aviation equipment based on risk interaction, characterized in that: The specific steps are as follows: Step 1: Collect and process various failure mode information; Through investigation, we obtain risk information from three aspects: FMEA data in the design stage, mechanical simulation test data, and failure damage data in the service stage. We pre-process the information and convert all of it into factor values ​​that have a positive impact on risk analysis. Then, we establish an incomplete information system for the risk of single structural components and measure the incomplete information system. The specific implementation process of preprocessing information is as follows: (1) For the FMEA data in the design phase, "severity level" and "probability level" are used as indicators for pre-analysis of component structure risk analysis. The language terms of "severity level" and "probability level" are converted into certain numbers between 0 and 1; according to the levels included in "severity" and "probability of occurrence", the corresponding number of certain numbers is set. The higher the level, the larger the corresponding certain value, as shown in the following table; (2) For the mechanical simulation test data information, "force magnitude" and "fracture coefficient" are used as indicators for in-process component structure risk analysis. To simplify the problem, the relationship between "force magnitude" and "fracture coefficient" is determined according to the following formula; In the formula, Q is the fragility coefficient, F is the force magnitude, E is the elastic modulus of the material, A is the cross-sectional area of ​​the material, L is the length of the material, q1 and q2 are constant coefficients, which depend on the specific experimental conditions and material properties; (3) Based on the information of the fault damage data in the service stage, "accident occurrence ratio" and "severity" are used as indicators for post-analysis of component structure risk analysis, and the language terms of "accident occurrence ratio" and "severity" are converted into definite numbers between 0 and 1; among which, the accident occurrence ratio is defined as the ratio of single component failure problems to total component failure problems, and the "severity" is determined according to the degree of harm of the failure problem. The more serious the degree of harm, the larger the corresponding definite value, as shown in the following table; (4) Establish an incomplete risk information system for monomer structural components. The specific implementation process is as follows: Let the decision object x i The set is: U = {x1, x2, ... x m }, m is the number of decision objects, and the attribute feature set of the evaluation decision object is: A = {a1, a2, ... a n }, n is the number of attribute features, and at the same time, decisions are made on information of k granularities, corresponding to k decision information systems, thus forming an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), where U k is a non-empty set of decision objects, C k is a non-empty finite set of conditional attributes, g k is a non-empty finite set of decision attributes, C k ∪g k Represents the set C k With the collection g k The union of k is a set of attribute values, f k For each object attribute value, through U k and C k ∪g k The Cartesian product of k The mapping function is obtained: f k :U k ×(C k ∪g k ) → A k ; The specific implementation process of measuring incomplete information systems is as follows: In an incomplete hybrid information system (U k ,C k ∪g k ,A k ,f k ), if the attribute value is a Boolean value, then the i-th decision object x at the k-th information granularity i The attribute value of the jth attribute Its vision Distance As shown below: If the attribute value is a certain number, the distance between the attribute value and the vision value As shown below: If the attribute is an interval number, the distance between the attribute value and the vision value As shown below: In the formula, is the attribute value interval, Respectively represent the lower and upper bounds of the attribute value interval; Step 2: Calculate the risk prospect utility value; According to the distance between the attribute value and the risk vision value, the risk loss matrix is ​​calculated. Then, considering the subjective factors of personnel in the risk analysis process, the risk loss matrix is ​​optimized using the prospect value function and converted into the actual expected utility value. That is, the larger the expected utility value, the smaller the corresponding risk value. The specific implementation process of the risk loss matrix is ​​as follows: According to the distance between the attribute value and the risk prospect value, the risk loss is defined, and the loss function corresponding to the three decision models under each attribute is established. Then the reference point is determined to obtain the profit and loss matrix: The specific implementation process of converting to actual expected utility value is as follows: Using the prospect value function V(x) in prospect theory, the profit and loss matrix is ​​mapped using the prospect value function to obtain the true risk expected utility value. Among them, V(x) represents the utility or value of the prospect in the value function, x represents the gain or loss of the prospect, α and β represent the gain and loss functions respectively, and are placed in the upper right corner to represent the power series, 0≤α,β≤1, θ represents the risk aversion coefficient, and when θ>1, it means that the decision maker is more sensitive to losses than gains; The utility value considering the risk preference of the decision maker is calculated. Taking into account the non-independence of attributes, the following λ fuzzy measure is used to fuse different attribute values: μ(A∪B)=μ(A)+μ(B)+λμ(A)μ(B) In the formula, μ(·) represents the attribute value, A and B represent two different attributes, A∪B is the union of attribute A and attribute B, λ is a measure between 0 and 1, and x i represents the i-th decision object, g(x i ) is x i The corresponding λ fuzzy measure; Step 3: Calculate the weights of the three conditional attributes of different granularities; The three data sources are called three-granularity data, where the design granularity is FMEA data, the test granularity is mechanical test data, and the reality granularity is accident and failure data. The attribute weight at each granularity is determined by calculating the information entropy method. Determine the attribute weight at each granularity. The specific implementation process is as follows: In terms of the degree of influence on the attribute weight, it is actually hoped that the more incomplete information there is, the smaller the attribute weight will be. Considering the vision value, it is converted into an actual utility value, and the attribute weight is solved using the value matrix. For the value matrix corresponding to the incomplete information system, if the value matrix of this attribute is more discrete, the attribute weight will be greater, otherwise, the attribute weight will be smaller. Therefore, the definition of information entropy is introduced, and the size of information entropy is used as a measure of attribute weight; The value entropy of the decision attribute is defined as: According to the entropy value of each attribute, the definition of attribute weight is given: In the formula, the attribute weight and K represents the maximum number of granularities or decision information systems, and J represents the number of attributes of the decision object; Step 4: Cluster similar risk attributes; Clustering of risks with similar attributes is done through risk utility deviation calculation, risk similarity calculation, and risk similarity classification, to simplify and accurately control risks. The specific implementation process of risk utility deviation calculation is as follows: First, the risk-related utility deviation of each component is calculated; the ratio of each risk actual utility value in each granular attribute to the maximum value in the attribute is defined as the risk-related utility deviation; in an incomplete mixed information system (U k ,C k ∪g k ,A k ,f k ), let Represents attribute value With vision value The relevant dispersion, the dispersion of the risk-related utility of each component Solve it by the following formula: In the formula, is the maximum value of the risk expected utility value among the j attributes of the i-th decision object at the k-th granularity; Based on the obtained risk-related utility deviations of each component and the weight values ​​of each attribute at each granularity, the risk utility deviations at different granularities are obtained. As shown in the following formula; The specific implementation process of risk similarity calculation is as follows: Based on the risk-related utility deviation, in an incomplete mixed information system (U k ,C k ∪g k ,A k ,f k ), let Csr k (x i ,x l ) represents the risk similarity between different components. Based on this, for any two decision objects x i ,x l ∈U, the risk similarity Csr between different decision objects at three granularities is obtained by the following cosine similarity calculation formula k (x i ,x l ); The specific implementation process of risk similarity classification is as follows: Decision object x i The δ-neighborhood of the same class is defined as: In the formula, δ is a similarity judgment value, 0≤δ≤1. When the similarity between two decision objects is greater than or equal to δ, it means that the decision object x i 、x l The risk level is the same; On the basis of similarity classification, the proportion of the number of components with the same risk level at each granularity to the total number of components is calculated according to the following probability measurement formula ρ(x i ); In the formula, ||U|| is the total number of decision objects, For the decision object x i The number of decision objects with the same risk level; Step 5: Calculate the utility matrix of each component under different state decisions; abstract the determination of key risk components of the structure into a decision problem, that is, for each analyzed monomer structural component, there are three decision classifications: one is that the component is considered to be a key risk component Second, it is considered that this component is not a critical risk component Third, it is considered that the component is at the boundary between the key risk component and the non-key risk component. Similarly, for the three decision categories, each component has three actual states, that is, the component is actually a key risk component S P , not actually a key risk component S N and uncertain S B ; Based on the acquired data, the decision utility matrix of each component in different states is calculated; Calculate the decision utility matrix of each component in different states. The specific implementation process is as follows: First, the probability calculation of decision division is to solve the correlation between each part and the three part decision classifications through the existing k granularity data; in (U k ,C k Ug k ,A k ,f k ) on the represents the decision state set, V gk ={A,H,R} represents the decision attribute g k The value range of Ω k The three states are defined as: In the formula, Indicates key risk components, A component that represents the boundary state between critical risks and non-critical risks. represents non-critical risk components; a, h, and r represent the attribute values ​​corresponding to three different decisions respectively; for three different decision object attribute states, there are three targeted decisions corresponding to them. When the decision object attribute is Accept the decision when the decision object attribute is When the decision object attribute is When decision making is made, the decision state set B = {b A ,b H ,b R }, where b A Accept the decision, b H represents delayed decision, b R Rejection of decision; In actual risk analysis, the actual accident problem data or new test accidents should be used as updated data to calculate the decision-making division probability. Since the safety risk analysis here does not involve updated data, the risk score not used in the design granularity FMEA is used as the updated data; according to the latest risk score data, different components are further divided into three levels of high risk, medium risk and low risk according to the predetermined score, so that x i Belong to the status The decision partition probability The calculation method is defined as: In the formula, ◇∈{A,H,R}, it means that the variable ◇ belongs to any variable among the variables A,H,R. k ,C k Ug k ,A k ,f k ), ρ(x i ) represents the ratio of the number of components with the same risk level to the number of all components at each granularity. The ratio of the total number of components in the high / medium / low risk level to which the component belongs to the total number of all components; Secondly, the associated utility matrix of each component in different states is calculated as: This matrix is ​​obtained by the following six associated loss functions; In the formula, Indicates that the component is actually a critical risk part, but the decision-maker considers that the component is at the boundary between the critical risk component and the non-critical risk component, resulting in losses; Indicates the loss caused by the fact that it is actually a key risk part, but the decision is that the part is not a key risk part; Indicates that the component is actually at the boundary state between the key risk component and the non-key risk component, but the decision-maker considers the component to be a key risk component and generates losses; The component is actually at the boundary between the critical risk component and the non-critical risk component, but the decision-maker considers the loss caused by the component as a non-critical risk component; Indicates that the component is actually a non-critical risk component, but the decision-maker considers the component to be a critical risk component; The component is actually a non-critical risk component, but the decision-maker believes that the component is at the boundary between the critical risk component and the non-critical risk component, resulting in losses; γ and μ are both loss magnification coefficients, is the risk utility dispersion at different granularities; Finally, combining the decision-making probability of each component in different states and the associated utility matrix of each component in different states, the utility matrix under different part decisions is calculated according to the following formula: In the formula, Indicates the utility value of accepting the decision when the decision-maker considers the component as a key risk component. It represents the utility value of the delayed decision when the decision-maker considers that the component is at the boundary state between the critical risk component and the non-critical risk component. It represents the utility value of rejecting the decision when the decision considers that the component is a non-critical risk component. Indicates the probability of considering this component as a key risk component, Indicates the probability of the component being considered as a boundary state between a critical risk component and a non-critical risk component, Indicates the probability of considering the component as a non-critical risk component. Step 6: Sort the risk of each component; divide the key parts under each granularity according to the set three-branch decision-making division rule; then use the given three granularities to assign weights, merge them, calculate the score of each key part, and finally get the risk of each component for ranking; The specific rules for dividing the three decisions are as follows: If satisfied: and have If satisfied: and have If satisfied: and have In the formula, A collection of key risk areas. is the boundary state set between key risk areas and non-key risk areas, For non-critical risk areas Calculate the score of each key part. The specific implementation process is as follows: The weights of multiple granularities in risk analysis are set as W = {W1, W2…W k }, where W1, W2…W k Respectively represent the weight coefficients of the 1st, 2nd…kth granularity; Set a signal function In the formula, Represents the decision object x at the kth granularity i The scores under different risk positions; Through the signal function, combined with the weights at different granularities, the final score R(x i ):

Citation Information

Cited By

  • Digitized workshop risk assessment method and system based on multivariate vector product

    CN120125026A