Improved fmeca analysis method for key assembly of rail vehicle based on faic model

Abstract

The application discloses an improved rail vehicle key assembly FMECA analysis method based on a FAIC model and belongs to the field of rail vehicle fault analysis. Firstly, in the pre-processing of input data, subjective and objective comprehensive weighting is performed on field expert evaluation information, so that the negative influence caused by excessive subjectivity of field experts is reduced. Secondly, in the establishment of a risk level evaluation model, the FAIC model is established by introducing the analytic hierarchy process and the CRITIC method, so that the credibility and practicability of the analysis result are improved. Compared with the traditional FMECA method, the method effectively improves the accuracy of the analysis result and can provide more credible argument support for the maintenance strategy of the rail vehicle key assembly in the actual maintenance process.

Description

Technical Field

[0001] This invention belongs to the field of railway vehicle fault analysis, specifically involving an improved railway vehicle FMECA analysis method based on the FAIC model. Background Technology

[0002] With the rapid development of rail vehicle technology, vehicle structures are becoming increasingly complex, and the requirements for the safety, reliability, and comfort of rail vehicles are also rising. The safe and reliable operation of rail vehicles is crucial for their normal operation. However, rail vehicles operate in complex environments with frequent vibrations and shocks, and drastic temperature changes. Their key assemblies will inevitably experience varying degrees of degradation, leading to malfunctions and potentially causing the rail vehicles to cease normal operation, resulting in serious safety accidents and economic losses. Therefore, research on the failure analysis of key rail vehicle assemblies has significant theoretical and practical implications.

[0003] However, the traditional FMECA method involves classifying and assigning values ​​to risk assessment indicators such as severity, occurrence, and detectability. These values ​​are provided by domain experts based on system or equipment operating data, their own knowledge, and their subjective judgment of the actual situation, thus introducing a degree of subjectivity. Secondly, experts in different domains vary in their professional level and familiarity with the problem. The traditional FMECA method simply smooths out the ratings from various experts, ignoring these differences and making the FMECA analysis highly subjective. Furthermore, the classification of severity, occurrence, and detectability is a fuzzy problem, but the traditional FMECA method does not perform fuzzification on the input data, leading to even more subjective results. Finally, the traditional FMECA risk priority number calculation does not consider the differences among the three indicators, resulting in a deviation between the calculated risk priority number and the actual situation. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, the present invention aims to provide an improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model. This method involves preprocessing the input data by comprehensively weighting the risk indicator evaluation information from domain experts using both subjective and objective methods, thereby reducing the adverse effects of excessive subjective factors. Furthermore, in calculating the risk priority number, the analytic hierarchy process (AHP) and the CRITIC method are introduced to establish the FAIC model, improving the scientific validity and reliability of the calculation results.

[0005] To achieve the above objectives, the solution of the present invention is:

[0006] An improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model is characterized by the following steps:

[0007] Step 1: Conduct FMEA analysis of key assemblies of rail vehicles, and introduce the "three-unit" decomposition method, namely "subsystem-functional component-failure mode", to summarize the failure of a key assembly of a rail vehicle into x functional components and y failure modes.

[0008] Step 2: Risk assessment index preprocessing. Invite domain experts to evaluate the severity, occurrence, and detectability of a failure mode of a key assembly of a rail vehicle. Evaluate the comprehensive capabilities of the domain experts based on their years of service and work experience, assigning subjective weights to them. Assign objective weights based on the quality of the evaluation information provided by the domain experts. In this way, determine the comprehensive weight vector of each expert's opinion, and then obtain the quantitative values ​​of severity S, occurrence O, and detectability D.

[0009] Step 3: Build a risk level assessment model. Based on the quantitative values ​​of severity S, occurrence O, and detectability D obtained in Step 2, establish a subjective weighting model using the analytic hierarchy process (AHP) combining triangular fuzzy numbers and the Grubbs test criterion, and establish an objective weighting model using the CRITIC method with the introduction of the coefficient of variation. Build the FAIC model and comprehensively assign weights to the three risk assessment indicators of severity S, occurrence O, and detectability D to obtain an improved risk level assessment model.

[0010] Furthermore, in step two, the expert opinions on the three risk assessment indicators of severity S, occurrence O, and detectability D are preprocessed. Since the comprehensive weighting solution process for the three risk assessment indicators of severity S, occurrence O, and detectability D is similar, only the comprehensive weighting solution process for severity is listed here.

[0011] First, based on the severity rating (S) given by domain experts, the severity rating information matrix SN is obtained:

[0012]

[0013] In the above formula, K represents the total number of experts, and N represents the total number of failure modes. This represents the subjective severity rating of the fault mode numbered n in a key assembly of a rail vehicle, given by the kth domain expert.

[0014] To obtain a scientifically reasonable quantitative value for the severity S, data fusion was performed on the evaluation information from domain experts:

[0015]

[0016] In the above formula, S n For the severity quantification of the failure mode numbered n, The overall weight of the k-th domain expert, including subjective weight. and objective weight

[0017] Subjective weight The calculation method is as follows:

[0018] The natural attributes of experts are evaluated from two dimensions: years of service and work experience. The subjective weight of the evaluation information is determined based on the natural attributes of experts in each field. The years of service refer to the length of time an expert has worked in the relevant field, denoted by the symbol 'a'. k1 The weight is denoted as w1; the work experience refers to the number of projects the domain expert has undertaken in the relevant field, denoted as a. k2 The weight it accounts for is denoted as w2;

[0019] First, the two evaluation indicators, years of service and work experience, were normalized:

[0020]

[0021] In the above formula, max{a k1} represents the maximum number of years of service, max{a k2} represents the maximum value of work experience; from this, the natural attribute evaluation matrix for experts in each field can be obtained:

[0022]

[0023] The indicator weight matrix is ​​denoted as:

[0024]

[0025] A comprehensive capability evaluation matrix for experts in various fields can be obtained:

[0026]

[0027] Therefore, the subjective weight of the k-th domain expert can be obtained:

[0028]

[0029]

[0030] The objective weight The calculation method is as follows:

[0031] The objective weights of domain experts are assigned based on the principle that the greater the conflict between individual and public opinions, the lower the data credibility.

[0032] First, the Euclidean distance is used to measure the distance between the severity S-level rating information of the fault mode numbered n given by the p-th expert and the q-th expert:

[0033]

[0034] The distance matrix for the severity evaluation index of the failure mode numbered n can be obtained:

[0035]

[0036] Transforming it to reduce it to the interval [0, 1], we can obtain the formula for calculating the similarity between the evaluation information given by the p-th expert and the q-th expert:

[0037]

[0038] Furthermore, a similarity matrix can be obtained among experts in all domains:

[0039]

[0040] The similarity matrix S among all domain experts n By summing each row, we can obtain the support level of the domain expert k for the severity index of the fault cause numbered n:

[0041]

[0042] After normalization, the objective weight of the k-th domain expert can be obtained:

[0043]

[0044]

[0045] The subjective weight of the k-th domain expert can be obtained from the above calculations. Objective weight of the kth domain expert Further, the comprehensive weight of the k-th domain expert can be obtained.

[0046]

[0047]

[0048] In the above formula, δ represents the weight preference number. Substituting the above formula into the formula for data fusion of domain expert evaluation information, and normalizing the fused evaluation value of severity S, we obtain the severity index for the fault mode numbered n:

[0049]

[0050] The process of calculating the combined weights of occurrence O and detectability D is similar to that of severity S, and will not be repeated here.

[0051] Furthermore, in step three, during the construction of the risk level assessment model, a subjective weighting model is established based on the analytic hierarchy process combining triangular fuzzy numbers and the Grubbs test criterion, an objective weighting model is established based on the CRITIC method with the introduction of the coefficient of variation, and a FAIC model is constructed. The three risk assessment indicators of severity (S), occurrence (O), and detectability (D) are comprehensively weighted, thereby obtaining an improved risk level assessment model.

[0052] The specific content of the subjective weighting model established by the analytic hierarchy process (AHP) combining triangular fuzzy numbers and the Grubbs test criterion is as follows:

[0053] First, a triangular fuzzy factor matrix is ​​constructed based on the quantitative scores of domain experts. This construction begins by establishing triangular fuzzy numbers and defining the membership function δ of the fuzzy number B. B (x):

[0054]

[0055] In the above formula, u is the upper limit of the value, m is the most likely value, and l is the lower limit of the value. The triangular fuzzy number B is denoted as B = [l, m, u].

[0056] Secondly, several domain experts were invited to conduct pairwise comparisons based on a relative importance metric scaling table to obtain the relative importance of the severity (S), occurrence (O), and detectability (D) indices. Based on this, a triangular fuzzy factor matrix composed of the relative importance information provided by the domain experts can be obtained:

[0057] F = (f ij ) M×M

[0058] In the above formula, M represents the number of risk assessment indicators, and f ij =[l ij ,m ij ,u ij [ ] represents the relative primary metric scale between the i-th and j-th indicators, and is a triangular fuzzy number, with m... ij Given a closed interval containing the median, the comprehensive triangular fuzzy number f can be obtained. ij for:

[0059]

[0060] In the above formula, M is the number of valid elements in matrix F;

[0061] Furthermore, the Grubbs test is introduced to remove obviously unreasonable data information, and the ratio of the absolute value of the residual of the element at the same position of a certain fuzzy factor in the triangular fuzzy matrix F to its standard deviation is calculated:

[0062]

[0063] In the above formula, b ij The elements to be detected include l ij ,m ij ,u ij , The average value of the elements to be detected is used as the basis for calculating G. Then, based on the required accuracy and data capacity, the corresponding confidence probability is selected. ij And compared with the corresponding standard critical value G in the Grubbs table. p In comparison, those with a G-value greater than G were eliminated. p Abnormal data;

[0064] The triangular fuzzy factor matrix F, which has passed the Grubbs test, is transformed as follows to obtain the triangular fuzzy evaluation factor matrix:

[0065]

[0066] In the above formula, The standard interest rate is used to reflect the degree of ambiguity in the evaluation information given by experts in the field.

[0067] Define matrix M as the median m of all triangular fuzziness factors in the triangular fuzziness factor matrix F. ij The matrix formed:

[0068]

[0069] The adjustment judgment matrix Q is obtained:

[0070] Q = M × R

[0071] The judgment matrix Q is transformed into a judgment matrix P = (p...) through a series of transformations. ij ) M×M And pij·pji=1;

[0072] By applying the compatibility matrix analysis method and transforming the judgment matrix P, the compatibility matrix R can be obtained:

[0073]

[0074] The root mean square (RMS) is used to determine the relative importance of each factor. The specific steps are as follows:

[0075] Multiplying the elements of each row of the compatibility matrix R, we get the vector T = (t i ) M×1 :

[0076]

[0077] Calculate the Mth power of each element in vector T to obtain vector U = (ui ) M×1 :

[0078]

[0079] Normalizing vector U yields vector W = (w i ) M×1 :

[0080]

[0081] We obtain W = (w i ) M×1 Since is the eigenvector of the compatibility matrix R, the subjective weights of the risk assessment indicators can be obtained:

[0082]

[0083] The specific content of establishing the objective weighting model based on the CRITIC method with introduced coefficient of variation is as follows:

[0084] First, an initial evaluation index data matrix X is constructed, which consists of risk evaluation index data for different failure modes:

[0085] X = (x nm ) N×M

[0086] In the above formula, x nm Let m be the value of the m-th risk assessment indicator for the fault mode numbered n, where n∈[1,N],m∈[1,M].

[0087] The Z-score method is used to standardize each element in the initial evaluation index data matrix to obtain the mean of the m-th index. and standard deviation

[0088]

[0089]

[0090] Based on the above formula, the standardized matrix can be obtained.

[0091]

[0092] Further, the correlation coefficient matrix G = (g kl ) M×M :

[0093]

[0094] In the above formula, g pqLet be the correlation coefficient between the p-th evaluation indicator and the q-th evaluation indicator. and The standardized matrix X is respectively * The average value of the p-th and q-th standardized evaluation index data;

[0095] Based on this, the independence coefficient θ of the m-th evaluation index can be obtained. m :

[0096]

[0097] The coefficient of variation J of the m-th evaluation index m Defined as the ratio of the standard deviation to the mean, i.e.:

[0098]

[0099] The comprehensive information content κ of the m-th evaluation index can be obtained. m :

[0100] κ m =J m ·θ m

[0101] The comprehensive information content κ of the m-th evaluation index m After normalization, the objective weight of the m-th evaluation indicator is obtained:

[0102]

[0103]

[0104] The above establishes subjective and objective weighting models sequentially. When applying the FAIC model to comprehensively weight risk assessment indicators, a preference factor is introduced to establish a subjective and objective weighting fusion model. The calculation formula for the fusion of subjective and objective weights is as follows:

[0105]

[0106] In the above formula, These are the comprehensive weight, subjective weight, and objective weight of the m-th evaluation indicator, respectively, and σ is the preference factor;

[0107] The preference factor optimization model is established as follows:

[0108]

[0109] The overall weight vector is obtained as follows:

[0110]

[0111] Based on the quantified values ​​of severity S, occurrence O, and detectability D obtained above, as well as the comprehensive weight vector, the improved risk level assessment model is constructed as follows:

[0112]

[0113] Compared with the prior art, the beneficial effects of the present invention are:

[0114] This invention presents an improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model. Addressing the shortcomings of traditional FMECA methods in analyzing key assemblies of rail vehicles, such as the strong subjectivity of input data and insufficient refinement of the RPN calculation model, the invention proposes corresponding improvements: In the preprocessing of input data, the evaluation information of domain experts is weighted both subjectively and objectively, reducing the negative impact of excessive subjectivity; in the construction of the risk level assessment model, the analytic hierarchy process (AHP) and the CRITIC method are introduced to build the FAIC model, improving the reliability and practicality of the analysis results and providing significant practical reference value. Attached Figure Description

[0115] Figure 1 The analysis process of the FMECA method based on the improved FAIC model is shown;

[0116] Figure 2 The algorithm for solving the quantitative values ​​of risk assessment indicators is shown.

[0117] Figure 3 The algorithm for solving the objective weights of risk assessment indicators is shown.

[0118] Figure 4 The diagram shows the hazard analysis of FMECA for improvements to key assembly functional components of rail vehicles;

[0119] Figure 5 The diagram shows the hazard analysis of the improvement of FMECA for key assembly subsystems of rail vehicles. Detailed Implementation

[0120] To enable those skilled in the art to better understand the present invention, the technical solution and beneficial effects of the present invention will be further explained below with reference to the accompanying drawings and specific examples in the embodiments of the present invention. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0121] This invention introduces a "three-unit" decomposition method, namely "subsystem-functional component-failure mode," using failure modes as the lowest agreed-upon level for FMEA analysis of key components in rail vehicles. The "three-unit" division method used in this invention's example of the high-speed train traction system categorizes failures into 15 functional components and 30 failure modes. The 15 functional components are numbered FE1-FE15 according to the order of the high-speed train traction system subsystems; the 30 failure modes are numbered &1-&30 according to the order of the failure modes. The FMEA table of the high-speed train traction system studied in this invention is shown in Table 1 below.

[0122] Table 1. FMEA Table for Traction System of High-Speed ​​EMU

[0123]

[0124]

[0125]

[0126] This invention takes the traction system of the CRH3 high-speed train as the research object. Ten experts in the field were invited to evaluate the various fault modes of the high-speed train traction system. For each fault mode, a unique evaluation level was selected based on its severity, occurrence, and detectability. The evaluation results of the fault modes of the high-speed train traction system are shown in Table 2-4.

[0127] Table 2 Severity evaluation information of the traction system failure modes of EMU trains

[0128]

[0129] Table 3. Occurrence Evaluation Information of Fault Modes in EMU Traction System

[0130]

[0131] Table 4. Detectability Evaluation Information of Fault Modes in EMU Traction System

[0132]

[0133] Based on Tables 2-4, the severity, occurrence, and detectability evaluation information matrices SN, ON, and DN are obtained respectively. Combining the specific calculation process of the risk assessment indicators described in Step 1, an algorithm based on MATLAB is used to solve for the quantitative values ​​of the risk assessment indicators for each fault mode of the high-speed train traction system. These risk assessment indicators include severity S, occurrence O, and detectability D. The specific solution process is shown in the attached figure. Figure 2 As shown.

[0134] Add the following indicator weight matrix w:

[0135] w = [w1 w2] T =[0.5 0.5] T

[0136] And the weight preference number δ = 0.5 between subjective and objective weights;

[0137] The final quantitative values ​​of risk assessment indicators for each fault mode of the high-speed train traction system are shown in Table 5.

[0138] Table 5. Quantitative values ​​of risk assessment indicators for failure modes of high-speed train traction systems.

[0139]

[0140] Next, the risk level assessment model was built. First, the FAIC model was used to subjectively assign weights to the risk assessment indicators in the risk level assessment model of the high-speed train traction system: 10 experts in the field were invited to compare the relative importance of the risk assessment indicators in pairs, and the relative importance of the indicators was evaluated based on the relative importance quantification scale table, which is shown in Table 6.

[0141] Table 6. Scale of Relative Importance Measurement

[0142]

[0143] Next, based on the formula... The Grubbs test was performed on the relative importance data provided by experts using the Grubbs table to remove obviously unreasonable data. The confidence level of the Grubbs table was set to 95%. The final triangular fuzzy factor judgment matrix F = (f ij ) 3×3 As shown below, where i,j∈[1,3], are the numbers of severity S, occurrence O, and detectability D, respectively:

[0144]

[0145] According to the following formula:

[0146]

[0147]

[0148] Q = M × R

[0149]

[0150] The compatibility matrix R = (r) of the risk assessment index can be calculated. ij ) 3×3 :

[0151]

[0152] Combine the following formula:

[0153]

[0154]

[0155]

[0156] The eigenvectors W = (w_i) of the compatibility matrix R can be calculated. i ) 3×1 :

[0157] W = [0.421 0.364 0.215] T

[0158] The subjective weight Φ of the risk assessment indicator 主 for:

[0159] Φ 主 =W T =[0.421 0.364 0.215]

[0160] Next, the FAIC model is used to objectively assign weights to the risk assessment indicators in the risk level assessment model of the high-speed train traction system. Combining the specific calculation process of the objective weights of the risk assessment indicators described in step two, an algorithm based on MATLAB is used to solve for the objective weights of the risk assessment indicators of the high-speed train traction system. These risk assessment indicators include severity S, occurrence O, and detectability D. The specific solution process is shown in the attached figure. Figure 3 As shown, the objective weights Φ of the risk assessment indicators are finally obtained. 客 for:

[0161] Φ 客 =[0.141 0.404 0.455]

[0162] Finally, when applying the FAIC model for comprehensive weighting, the first step is to use the formula...

[0163]

[0164] A preference factor optimization model was built and solved, yielding a preference factor σ = 0.486;

[0165] Based on this, the comprehensive weight Φ of the risk assessment indicators is obtained:

[0166] Φ = [0.385 0.338 0.277]

[0167] Finally, combining Table 5 and the formula The risk priority number corresponding to each failure mode can be obtained, that is, the severity of the failure mode. The specific results are shown in Table 7. The severity analysis of FMECA for the improvement of functional components of the high-speed train traction system is attached. Figure 4 As shown in the attached figure, the hazard analysis of the FMECA of improving the traction system subsystem of high-speed trains is as follows. Figure 5 As shown.

[0168] Table 7. Hazard level of high-speed train traction system based on improved FMECA analysis.

[0169]

[0170] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. An improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model, characterized in that... Includes the following steps: Step 1: Conduct FMEA analysis of key assemblies of rail vehicles, and introduce the "three-unit" decomposition method, namely "subsystem-functional component-failure mode", to summarize the failures of key assemblies of rail vehicles into x functional components and y failure modes. Step 2: Risk assessment index preprocessing. Inviting domain experts to evaluate the severity, occurrence, and detectability of the failure modes of key components of rail vehicles. The comprehensive capabilities of the domain experts are evaluated based on their years of service and work experience, and subjective weights are assigned. Objective weights are assigned based on the quality of the evaluation information provided by the domain experts. This determines the comprehensive weight vector of each expert's opinion, and then the quantitative values ​​of severity S, occurrence O, and detectability D are obtained. Step 3: Build a risk level assessment model. Based on the quantitative values ​​of severity S, occurrence O, and detectability D obtained in Step 2, establish a subjective weighting model using the analytic hierarchy process (AHP) combining triangular fuzzy numbers and the Grubbs test criterion, and establish an objective weighting model using the CRITIC method with the introduction of the coefficient of variation. Build the FAIC model and comprehensively assign weights to the three risk assessment indicators of severity S, occurrence O, and detectability D to obtain an improved risk level assessment model.

2. The improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model according to claim 1, characterized in that: In step two, the expert opinions on the three risk assessment indicators of severity S, occurrence O and detectability D are preprocessed. Since the comprehensive weighting solution process of the three risk assessment indicators of severity S, occurrence O and detectability D is similar, this claim only lists the comprehensive weighting solution process of severity. First, based on the severity rating (S) given by domain experts, the severity rating information matrix SN is obtained: In the above formula, K represents the total number of experts, and N represents the total number of failure modes. This represents the subjective severity rating of the failure mode numbered n in the key assembly of the rail vehicle, given by the kth domain expert. To obtain a scientifically reasonable quantitative value for the severity S, data fusion was performed on the evaluation information from domain experts: In the above formula, S n For the severity quantification of the failure mode numbered n, The overall weight of the k-th domain expert, including subjective weight. and objective weight Subjective weight The calculation method is as follows: The natural attributes of experts are evaluated from two dimensions: years of service and work experience. The subjective weight of the evaluation information is determined based on the natural attributes of experts in each field. The years of service refer to the length of time an expert has worked in the relevant field, denoted by the symbol 'a'. k1 The weight is denoted as w1; the work experience refers to the number of projects the domain expert has undertaken in the relevant field, denoted as a. k2 The weight it accounts for is denoted as w2; First, the two evaluation indicators, years of service and work experience, were normalized: In the above formula, max{a k1 } represents the maximum number of years of service, max{a k2 } represents the maximum value of work experience; This allows us to obtain the natural attribute evaluation matrix for experts in various fields: The indicator weight matrix is ​​denoted as: A comprehensive capability evaluation matrix for experts in various fields can be obtained: Therefore, the subjective weight of the k-th domain expert can be obtained: The objective weight The calculation method is as follows: The objective weights of domain experts are assigned based on the principle that the greater the conflict between individual and public opinions, the lower the data credibility. First, the Euclidean distance is used to measure the distance between the severity S-level rating information of the fault mode numbered n given by the p-th expert and the q-th expert: The distance matrix for the severity evaluation index of the failure mode numbered n can be obtained: Transforming it to reduce it to the interval [0, 1], we can obtain the formula for calculating the similarity between the evaluation information given by the p-th expert and the q-th expert: Furthermore, a similarity matrix can be obtained among experts in all domains: The similarity matrix S among all domain experts n By summing each row, we can obtain the support level of the domain expert k for the severity index of the fault cause numbered n: After normalization, the objective weight of the k-th domain expert can be obtained: The subjective weight of the k-th domain expert can be obtained from the above calculations. Objective weight of the kth domain expert Further, the comprehensive weight of the k-th domain expert can be obtained. In the above formula, δ represents the weight preference number. Substituting the above formula into the formula for data fusion of domain expert evaluation information, and normalizing the fused evaluation value of severity S, we obtain the severity index for the fault mode numbered n: The process of calculating the combined weights of occurrence O and detectability D is similar to that of severity S, and will not be repeated here.

3. The improved FMECA analysis method for key assemblies of rail vehicles based on the FAIC model according to claim 1, characterized in that: In step three, during the construction of the risk level assessment model, a subjective weighting model is established based on the analytic hierarchy process combining triangular fuzzy numbers and Grubbs' test criterion, an objective weighting model is established based on the CRITIC method with the introduction of the coefficient of variation, and a FAIC model is built. The three risk assessment indicators of severity (S), occurrence (O), and detectability (D) are comprehensively weighted to obtain an improved risk level assessment model. The specific content of the subjective weighting model established by the analytic hierarchy process (AHP) combining triangular fuzzy numbers and the Grubbs test criterion is as follows: First, a triangular fuzzy factor matrix is ​​constructed based on the quantitative scores of domain experts. This construction begins by establishing triangular fuzzy numbers and defining the membership function δ of the fuzzy number B. B (x): In the above formula, u is the upper limit of the value, m is the most likely value, and l is the lower limit of the value. The triangular fuzzy number B is denoted as B = [l, m, u]. Secondly, several domain experts were invited to conduct pairwise comparisons based on a relative importance metric scaling table to obtain the relative importance of the severity (S), occurrence (O), and detectability (D) indices. Based on this, a triangular fuzzy factor matrix composed of the relative importance information provided by the domain experts can be obtained: F=(f ij ) M×M In the above formula, M represents the number of risk assessment indicators, and f ij =[l ij ,m ij ,u ij [ ] represents the relative primary metric scale between the i-th and j-th indicators, and is a triangular fuzzy number, with m... ij Given a closed interval containing the median, the comprehensive triangular fuzzy number f can be obtained. ij for: In the above formula, M is the number of valid elements in matrix F; Furthermore, the Grubbs test is introduced to remove obviously unreasonable data information, and the ratio of the absolute value of the residual of the element at the same position of a certain fuzzy factor in the triangular fuzzy matrix F to its standard deviation is calculated: In the above formula, b ij The elements to be detected include l ij ,m ij ,u ij , The average value of the elements to be detected is used as the basis for calculating G. Then, based on the required accuracy and data capacity, the corresponding confidence probability is selected. ij And compared with the corresponding standard critical value G in the Grubbs table. p In comparison, those with a G-value greater than G were eliminated. p Abnormal data; The triangular fuzzy factor matrix F, which has passed the Grubbs test, is transformed as follows to obtain the triangular fuzzy evaluation factor matrix: In the above formula, The standard interest rate is used to reflect the degree of ambiguity in the evaluation information given by experts in the field. Define matrix M as the median m of all triangular fuzziness factors in the triangular fuzziness factor matrix F. ij The matrix formed: The adjustment judgment matrix Q is obtained: Q = M × R The judgment matrix Q is transformed into a judgment matrix P = (p...) through a series of transformations. ij ) M×M And p ij ·p ji =1; By applying the compatibility matrix analysis method and transforming the judgment matrix P, the compatibility matrix R can be obtained: The root mean square (RMS) is used to determine the relative importance of each factor. The specific steps are as follows: Multiplying the elements of each row of the compatibility matrix R, we get the vector T = (t i ) M×1 : Calculate the Mth power of each element in vector T to obtain vector U = (u i ) M×1 : Normalizing vector U yields vector W = (w i ) M×1 : We obtain W = (w i ) M×1 Since is the eigenvector of the compatibility matrix R, the subjective weights of the risk assessment indicators can be obtained: The specific content of establishing the objective weighting model based on the CRITIC method with introduced coefficient of variation is as follows: First, an initial evaluation index data matrix is ​​constructed. The initial evaluation index data matrix X consists of various risk evaluation index data for different failure modes: X=(x nm ) N×M In the above formula, x nm Let m be the value of the m-th risk assessment indicator for the fault mode numbered n, where n∈[1,N],m∈[1,M]. The Z-score method is used to standardize each element in the initial evaluation index data matrix to obtain the mean of the m-th index. and standard deviation Based on the above formula, the standardized matrix can be obtained. Further, the correlation coefficient matrix G = (g kl ) M×M : In the above formula, g pq Let be the correlation coefficient between the p-th evaluation indicator and the q-th evaluation indicator. and The standardized matrix X is respectively * The average value of the p-th and q-th standardized evaluation index data; Based on this, the independence coefficient θ of the m-th evaluation index can be obtained. m : The coefficient of variation J of the m-th evaluation index m Defined as the ratio of the standard deviation to the mean, i.e.: The comprehensive information content κ of the m-th evaluation index can be obtained. m : k m =J m ·i m The comprehensive information content κ of the m-th evaluation index m After normalization, the objective weight of the m-th evaluation indicator is obtained: The above establishes subjective and objective weighting models sequentially. When applying the FAIC model to comprehensively weight risk assessment indicators, a preference factor is introduced to establish a subjective and objective weighting fusion model. The calculation formula for the fusion of subjective and objective weights is as follows: In the above formula, These are the comprehensive weight, subjective weight, and objective weight of the m-th evaluation indicator, respectively, and σ is the preference factor; The preference factor optimization model is established as follows: The overall weight vector is obtained as follows: Based on the quantified values ​​of severity S, occurrence O, and detectability D obtained in claim 2, and the comprehensive weight vector obtained above, an improved risk level assessment model is constructed as follows:

Images