Failure model risk analysis device and method for improved FMEA

Abstract

The invention provides an improved FMEA failure model risk analysis device and method, and the device is characterized in that the device comprises an initial evaluation matrix generation module which is used for calculating a failure mode fuzzy initial evaluation matrix; the risk factor comprehensive weight calculation module is used for calculating the comprehensive weight corresponding to each risk factor; the DEA evaluation score calculation module is used for calculating to obtain a cross efficiency value and a DEA evaluation score corresponding to each failure mode; the comprehensive evaluation calculation module is used for calculating a Fuzzy Dombi evaluation value and a comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, the comprehensive weight, the cross efficiency value and the DEA evaluation score; and the risk analysis module is used for obtaining a risk sorting result and a risk rating of each failure mode as a risk analysis result. In a word, the method can generate a more reliable and accurate failure mode risk analysis result.

Description

Technical Field

[0001] The present invention relates to the technical field of reliability analysis methods, and in particular to a failure model risk analysis device and method for improving FMEA. Background Art

[0002] Failure Mode and Effects Analysis (FMEA) is a forward-looking, systematic reliability analysis and safety assessment technology. Failure Mode and Effects Analysis is used to analyze technical risks and causes of failures, evaluate the impact of different failure modes and allocate limited resources to reduce risks, improve or eliminate failure modes, and help the system improve safety and reliability. Risk factors are generally used to describe failure mode evaluation dimensional indicators in order to assess risks and rank failure modes according to risk priority number (RPN) values. This step requires experts to have an in-depth understanding of the RPN calculation method so that they can accurately calculate and rank the RPN value.

[0003] However, existing failure mode analysis still has problems such as experts' scores are easily influenced by their own preferences, resulting in large deviations and fluctuations in the evaluation results, and the values ​​obtained by quantitative calculations cannot fully reflect the advantages and disadvantages of the failure mode, resulting in unreliable results. Summary of the invention

[0004] The present invention is made to solve the above problems, and aims to provide a failure model risk analysis device and method for improving FMEA.

[0005] The present invention provides a failure model risk analysis device for improving FMEA, which is used for obtaining risk analysis results of all failure modes according to a failure mode scoring matrix formed by scores of each expert on each risk factor of each failure mode, a risk factor weight evaluation matrix formed by scores of each expert on each risk factor, and expert trust scores of all experts. The device has the following characteristics, including: an initial evaluation matrix generation module, which is used for calculating a fuzzy initial evaluation matrix of failure modes according to expert trust scores and the failure mode scoring matrix; a risk factor comprehensive weight calculation module, which is used for calculating the comprehensive weights corresponding to each risk factor according to the risk factor weight evaluation matrix and the fuzzy initial evaluation matrix of failure modes; a data envelopment analysis (DEA) evaluation score calculation module, which stores a preset adversarial DEA model, which is used for calculating the cross efficiency value and DEA evaluation score corresponding to each failure mode according to the adversarial DEA model according to the failure mode fuzzy initial evaluation matrix; and a comprehensive evaluation calculation module, which is used for calculating the fuzzy initial evaluation matrix, comprehensive weights, cross efficiency values ​​and DEA evaluation scores corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix. Dombi evaluation value and comprehensive evaluation value; risk analysis module, which stores the adversarial structure interpretation model and adaptive resonance neural network, is used to perform risk sorting and cluster analysis on all failure modes according to the comprehensive evaluation value, DEA evaluation score and Fuzzy Dombi evaluation value, and obtain the risk sorting result and risk rating of each failure mode as the risk analysis result.

[0006] In the improved FMEA failure model risk analysis device provided by the present invention, it can also have the following characteristics: wherein, the initial evaluation matrix generation module includes: a preference relationship matrix generation unit, which stores a preset correction formula, and is used to calculate the trust scores of all experts according to the correction formula to obtain a preference relationship matrix containing preference information corresponding to each other among experts; a feature preference relationship matrix generation unit, which is used to calculate the feature preference relationship matrix according to the preference relationship matrix; a consistency judgment unit, which stores preset judgment conditions and adjustment formulas, and is used to judge whether the feature preference relationship matrix meets the preset judgment conditions. If so, the feature preference relationship matrix is ​​used as a consistent preference relationship matrix. If not, the feature preference relationship matrix is ​​adjusted according to the adjustment formula, and the consistency judgment unit is re-executed; a weighted average unit, which is used to calculate the elements corresponding to each expert in the consistent preference relationship matrix respectively, and obtain the comprehensive preference information corresponding to each expert; an expert weight calculation unit, which is used to normalize all comprehensive preference information to obtain the expert weight corresponding to each expert; an aggregation unit, which stores an evaluation semantic set, and is used to convert the failure mode scoring matrix into a failure mode fuzzy initial evaluation matrix through the Fuzzy Dombi (FD) algorithm according to the expert weight and the evaluation semantic set.

[0007] The improved FMEA failure model risk analysis device provided by the present invention may also have the following features: wherein, in the feature preference relationship matrix generation unit, the preference relationship matrix middle The calculation expression is: In the formula is the preference relationship matrix The element in row i and column j in p i+s,i+s+1 is the preference relationship matrix P = (p ij ) n×n In the consistency judgment unit, the expression of the preset judgment condition is: CI(P (t) )≤δ0, where P (t) is the current feature preference relationship matrix, CI(P (t) ) is the feature preference relationship matrix P (t) The consistency index, δ0 ​​is the consistency judgment threshold, and the expression of the adjustment formula is: In the formula is the element in the i-th row and j-th column of the adjusted feature preference relationship matrix, is the element in the i-th row and j-th column of the preference relationship matrix before adjustment, is the element in the i-th row and j-th column of the feature preference relationship matrix before adjustment, θ is the weight coefficient, N is the set of all expert pairs, and in the weighted average unit, the calculation expression of the comprehensive preference information is: Where p i is the comprehensive preference information of the ith expert, n is the total number of experts, p ij is the element in the i-th row and j-th column of the consistent preference relationship matrix.

[0008] The improved FMEA failure model risk analysis device provided by the present invention may also have the following features: wherein the risk factor comprehensive weight calculation module includes: an interval rough number matrix generation unit, which stores an interval semantic evaluation set and is used to convert the risk factor weight evaluation matrix into an interval rough number matrix; a subjective weight calculation unit, which stores an interval rough number expectation formula and is used to calculate the interval rough number matrix according to the interval rough number expectation formula to obtain the subjective weight corresponding to each risk factor; a defuzzification unit, which is used to generate the subjective weight corresponding to each risk factor according to the PV m The weighting method defuzzifies the failure mode fuzzy initial evaluation matrix to obtain a defuzzified evaluation matrix; the objective weight calculation unit is used to calculate the objective weight corresponding to each risk factor according to the defuzzified evaluation matrix; the comprehensive weight calculation unit is used to calculate the corresponding comprehensive weight for each risk factor according to the corresponding subjective weight and objective weight.

[0009] The improved FMEA failure model risk analysis device provided by the present invention may also have the following features: wherein, in the defuzzification unit, the defuzzification expression is: In the formula is the element in the i-th row and j-th column of the defuzzified evaluation matrix, is the element in the i-th row and j-th column of the fuzzy initial judgment matrix of the failure mode, a ij l 、a ij m and a ij u They are the left value, middle value and right value of the triangular fuzzy number corresponding to the evaluation semantics of the i-th row and j-th column. For each risk factor, the adjustment coefficient of the subjective weight and the adjustment coefficient of the objective weight are both set to 0.5 and combined and weighted to obtain the comprehensive weight.

[0010] In the improved FMEA failure model risk analysis device provided by the present invention, the following features may also be provided: wherein the mathematical model of the adversarial DEA model is as follows: objective function: Constraints: Where x id is the i-th input variable of the d-th failure mode, v id The input variable x id The associated weight, y rd is the rth output variable of the dth failure mode, u rd is the output variable y rd The relevant weights, is the self-assessment efficiency value of the d-th failure mode, s is the output dimension, m is the input dimension, the risk factors include severity S, occurrence O, detection D, repair difficulty R and repair cost C, the input variables are severity S, occurrence O and detection D, the output variables are repair difficulty R and repair cost C, and the calculation expression of DEA evaluation score is: In the formula is the DEA evaluation score corresponding to the ith failure mode, x i is the cross efficiency value of the ith failure mode obtained according to the adversarial DEA model.

[0011] In the improved FMEA failure model risk analysis device provided by the present invention, it can also have the following characteristics: wherein, the comprehensive evaluation calculation module includes: a support matrix generation unit, which is used to normalize the failure mode fuzzy initial evaluation matrix to obtain the failure mode fuzzy decision support matrix; an FD score calculation unit, which is used to aggregate the failure mode fuzzy decision support matrix through the Fuzzy Dombi method according to the comprehensive weight and cross efficiency value, and calculate the Fuzzy Dombi evaluation value corresponding to each failure mode; a comprehensive evaluation calculation unit, which is used to calculate the comprehensive evaluation value corresponding to each failure mode according to the Fuzzy Dombi evaluation value and the DEA evaluation score.

[0012] The improved FMEA failure model risk analysis device provided by the present invention may also have the following features: wherein, in the support matrix generation unit, the normalized calculation expression is:

[0013] In the formula is the element in the i-th row and j-th column of the failure mode fuzzy decision support matrix, a ij is the element in the i-th row and j-th column of the fuzzy initial evaluation matrix of the failure mode. The severity S, occurrence O and detection D in the risk factors are positive indicators. The repair difficulty R and repair cost C in the risk factors are negative indicators. In the FD score calculation unit, the weight of the arithmetic mean operator and the weight of the geometric mean operator of the Fuzzy Dombi method are set to 0.8 and 0.2 respectively. The comprehensive weight and cross efficiency value are used as the operator fusion parameters of the Fuzzy Dombi method. In the comprehensive evaluation calculation unit, for each failure mode, the weight of the corresponding Fuzzy Dombi evaluation value is set to 0.8, and the weight of the corresponding DEA evaluation score is set to 0.2. The corresponding comprehensive evaluation value is obtained by weighted summation.

[0014] In the improved FMEA failure model risk analysis device provided by the present invention, it can also have the following characteristics: wherein, the risk analysis module includes: a risk sorting unit, which is used to sort the failure modes according to the comprehensive evaluation value; a risk rating unit, which stores the adversarial structure explanation model and the adaptive resonance neural network, and is used to input the Fuzzy Dombi evaluation value and DEA evaluation score corresponding to each failure mode as characteristic values ​​into the adversarial structure explanation model and the adaptive resonance neural network to obtain multiple clusters corresponding to different risk levels, the smaller the comprehensive evaluation value, the smaller the corresponding failure mode sequence number, the greater the risk corresponding to the failure mode, and the risk rating of the failure mode in the cluster is the risk level corresponding to the cluster.

[0015] The present invention also provides a failure model risk analysis method for improving FMEA, which is used to obtain risk analysis results of all failure modes according to a failure mode scoring matrix composed of scores of each expert on each risk factor of each failure mode, a risk factor weight evaluation matrix composed of scores of each expert on each risk factor, and expert trust scores of all experts, and has the following characteristics, including: step S1, calculating a failure mode fuzzy initial evaluation matrix according to the expert trust scores and the failure mode scoring matrix; step S2, calculating the comprehensive weights corresponding to each risk factor according to the risk factor weight evaluation matrix and the failure mode fuzzy initial evaluation matrix; step S3, calculating the cross efficiency value and DEA evaluation score corresponding to each failure mode through an adversarial DEA model according to the failure mode fuzzy initial evaluation matrix; step S4, calculating the Fuzzy Dombi evaluation value and the comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, the comprehensive weight, the cross efficiency value and the DEA evaluation score; step S5, calculating the comprehensive evaluation value, the DEA evaluation score and the Fuzzy Dombi evaluation score according to the comprehensive evaluation value, the DEA evaluation score and the Fuzzy Dombi evaluation score. The Dombi evaluation value, combined with the adversarial structural interpretation model and the adaptive resonance neural network, performs risk ranking and cluster analysis on all failure modes, and obtains the risk ranking results and risk ratings of each failure mode as the risk analysis results.

[0016] Functions and Effects of the Invention

[0017] According to the improved FMEA failure model risk analysis device and method involved in the present invention, through the initial evaluation matrix generation module, the expert weight is calculated based on the preference relationship, so as to fully consider the difference in the expert's preference strength, reduce the influence of this factor on the expert's decision, and make the decision more real and reliable; through the risk factor comprehensive weight calculation module, the interval rough number is used to obtain the subjective weight according to the risk factor weight evaluation matrix, and the logarithmic percentage change driven weighting method is used to obtain the objective weight, and then the comprehensive weight is obtained; through the comprehensive evaluation calculation module and the risk analysis module, the adversarial data envelopment analysis and the Fuzzy Dombi method are used to sort the failure modes, and the advantages and disadvantages of the failure modes are evaluated more comprehensively, and the results are more reliable, and the adaptive resonance neural network is used for cluster analysis. Therefore, the improved FMEA failure model risk analysis device and method of the present invention can generate more reliable and accurate failure mode risk analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic diagram of the scores of various risk factors of various failure modes by various experts in an embodiment of the present invention;

[0019] Figure 2 is a block diagram of a failure model risk analysis device in an embodiment of the present invention;

[0020] Figure 3 is a schematic diagram of fuzzy initial evaluation of failure modes in an embodiment of the present invention;

[0021] Figure 4 is a schematic diagram of the content of the failure mode fuzzy decision support matrix in an embodiment of the present invention;

[0022] Figure 5 is a schematic diagram of a cluster in an embodiment of the present invention;

[0023] Figure 6 It is a flow chart of a failure model risk analysis method for improving FMEA in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following embodiments are combined with the accompanying drawings to specifically illustrate the failure model risk analysis device and method of the present invention for improving FMEA.

[0025] In this embodiment, an improved FMEA failure model risk analysis device is provided, hereinafter referred to as the failure model risk analysis device, which is used to obtain risk analysis results of all failure modes based on a failure mode scoring matrix formed by scores of each expert on each risk factor of each failure mode, a risk factor weight evaluation matrix formed by scores of each expert on each risk factor, and expert trust scores of all experts.

[0026] In this embodiment, the failure model risk analysis device of this embodiment is explained by performing failure mode analysis on the ship operation and maintenance digital twin system.

[0027] In this embodiment, the experts include university scholars, digital twin system technical managers, senior information project managers, and digital twin system implementation consultants, respectively denoted as expert 1, expert 2, expert 3, and expert 4.

[0028] In this embodiment, through literature review, on-site investigation and expert interviews, a total of 24 failure modes were obtained, as shown in the following table:

[0029] Failure Mode Number Failure Mode FM1 Unreasonable route planning FM2 Route optimization errors FM3 Route avoidance error FM4 Emission Control Area (ECA) identification was unsuccessful FM5 Main propulsion system and shafting monitoring failure FM6 Auxiliary power system monitoring failure FM7 Liquid level telemetry monitoring error FM8 Power system monitoring failure FM9 Steam system monitoring failure FM10 Health assessment gone wrong FM11 Cause list error FM12 Wrong maintenance strategy selection FM13 Assisted suggestion error FM14 Imperfect fuel consumption optimization FM15 Power consumption monitoring failure FM16 Emissions management failure FM17 Error in warehouse management FM18 Untimely planned maintenance FM19 Procurement information management failure FM20 Cost control failure FM21 Data collection failure FM22 Data storage loss FM23 Analysis and alarm are not timely FM24 Performance evaluation errors

[0030] The first column in the table above is the number of each failure mode, and the second column is the specific content of the failure mode corresponding to each number. For example, the cell in the second row and second column indicates that the specific content of the failure mode numbered FM1 is "unreasonable route planning".

[0031] In this embodiment, the risk factors include severity S, occurrence O, detection D, repair difficulty R and repair cost C. Experts use the five-scale method to score each risk factor of each failure mode through the evaluation standard table to obtain the corresponding evaluation score. The evaluation standard table is as follows:

[0032]

[0033]

[0034] The first column in the table above shows the evaluation scores, and the second to sixth columns respectively show the descriptions of the situations corresponding to the evaluation scores under the repair difficulty R, repair cost C, severity S, occurrence O, and detection D. For example, the cell in the second row and second column indicates that the evaluation score 1 in the repair difficulty R means that the operator can repair the problem by himself.

[0035] Figure 1 It is a schematic diagram of the scores of various experts on various risk factors of various failure modes in an embodiment of the present invention.

[0036] like Figure 1 As shown in Figure 1, each cell in the failure mode scoring matrix corresponds to an expert's evaluation score for a risk factor of a failure mode.

[0037] In this embodiment, each risk factor is scored according to the risk factor importance evaluation standard table, and the risk factor importance evaluation standard table is as follows:

[0038] Importance score Score meaning 1 Almost insignificant 2 Slightly affected 3，4 A lower degree of impact 5，6 More important 7 Big impact 8 Play a decisive role 9 Crucial

[0039] The first column in the formula is the importance score, and the second column is the meaning corresponding to the importance score.

[0040] Based on the above risk factor importance evaluation standard table, the experts' scores for each risk factor are shown in the following table:

[0041] expert Severity S Occurrence rate Detection degree D Repair difficulty R Repair cost C Expert 1 9 8 6 5 7 Expert 2 8 6 7 7 5 Expert 3 9 7 8 7 5 Expert 4 9 8 7 6 5

[0042] The first column in the above table is the scores of each expert, and the second to sixth columns are the scores of severity S, occurrence O, detection D, repair difficulty R and repair cost C respectively. Then, a risk factor weight evaluation matrix is ​​constructed based on the above table. Each cell in the matrix corresponds to the score of a certain expert on a certain risk factor.

[0043] In this embodiment, each expert scores the trustworthiness of other experts using the expert trustworthiness evaluation standard table, which is as follows:

[0044] Trust Score Trust level 1 Very low 2 Low 3 medium 4 high 5 Very high

[0045] The first column in the above table is the trust score, and the second column is the trust level corresponding to the trust score.

[0046] According to the above expert trust evaluation standard table, the expert trust scores of each expert are shown in the following table:

[0047] expert Expert 1 Expert 2 Expert 3 Expert 4 Expert 1 - 4 3 5 Expert 2 4 - 3 5 Expert 3 3 4 - 5 Expert 4 3 4 3 -

[0048] The first column in the table above is the review experts, and the second to fifth columns are the expert trust scores of each review expert on expert 1 to expert 4. For example, the cell in the third column of the second row indicates that expert 1's expert trust score for expert 2 is 4.

[0049] Figure 2 It is a block diagram of a failure model risk analysis device in an embodiment of the present invention.

[0050] like Figure 2 As shown, the failure model risk analysis device 100 includes an initial evaluation matrix generation module 10, a risk factor comprehensive weight calculation module 20, a DEA evaluation score calculation module 30, a comprehensive evaluation calculation module 40, a risk analysis module 50 and a control module 60 for controlling the above modules.

[0051] The initial evaluation matrix generation module 10 is used to calculate the failure mode fuzzy initial evaluation matrix according to the expert trust degree score and the failure mode score matrix.

[0052] The initial evaluation matrix generation module 10 includes a preference relationship matrix generation unit 101 , a feature preference relationship matrix generation unit 102 , a consistency judgment unit 103 , a weighted average unit 104 , an expert weight calculation unit 105 and an aggregation unit 106 .

[0053] The preference relationship matrix generating unit 101 stores a preset correction formula, which is used to calculate the trust scores of all experts according to the correction formula to obtain a preference relationship matrix containing the corresponding preference information between two experts.

[0054] The preference information of each expert in this embodiment is shown in the following table:

[0055] Expert 1 Expert 2 Expert 3 Expert 4 Expert 1 0.5000 0.4048 0.5667 0.3148 Expert 2 0.6176 0.5000 0.7000 0.3889 Expert 3 0.4412 0.3571 0.5000 0.2778 Expert 4 0.7941 0.6429 0.9000 0.5000

[0056] The first column in the table above is the review expert, and the second to fifth columns are the preference information of each review expert for experts 1 to 4, i.e., the reviewed experts. For example, the cell in the third column of the second row indicates that the preference information of expert 1 for expert 2 is 0.4048. In this embodiment, the size of the preference information is used to reflect the degree of trust between experts. The larger the preference information, the higher the degree of trust of the review expert in the reviewed expert.

[0057] The feature preference relationship matrix generating unit 102 is used to calculate the feature preference relationship matrix according to the preference relationship matrix.

[0058] Among them, in the feature preference relationship matrix generating unit 102, the preference relationship matrix middle The calculation expression is:

[0059]

[0060] In the formula is the preference relationship matrix The element in row i and column j in p i+s,i+s+1 is the preference relationship matrix P = (p ij ) n×n The element at the i+sth row and i+s+1th column in .

[0061] In this embodiment, the preference relationship matrix Satisfies additive consistency, that is

[0062] The consistency judgment unit 103 stores preset judgment conditions and adjustment formulas, which are used to judge whether the feature preference relationship matrix meets the preset judgment conditions. If so, the feature preference relationship matrix is ​​used as a consistent preference relationship matrix. If not, the feature preference relationship matrix is ​​adjusted according to the adjustment formula and the consistency judgment unit 103 is re-executed.

[0063] In the consistency judgment unit 103, the expression of the preset judgment condition is:

[0064] CI(P (t) )≥δ0,

[0065] Where P (t) is the current feature preference relationship matrix, CI(P (t) ) is the feature preference relationship matrix P (t) The consistency index is δ0, and δ0 is the consistency judgment threshold. In this embodiment, δ0 = 0.01. In this embodiment, CI (P (t) ) is used to judge the inconsistency of weighting elements at the same level, that is, to judge the weight of the same expert trust evaluation. If CI(P (t) )≤δ0, the matrix is ​​considered to have high consistency.

[0066] In this embodiment, a weighted update strategy is used to gradually make the adjustment value consistent, while not overly relying on a single reference value, thereby avoiding excessive oscillation or instability of the adjustment value. Therefore, the expression of the adjustment formula is:

[0067]

[0068] In the formula is the element in the i-th row and j-th column of the adjusted feature preference relationship matrix, is the element in the i-th row and j-th column of the preference relationship matrix before adjustment, is the element in the i-th row and j-th column of the feature preference relationship matrix before adjustment, θ is the weight coefficient, and N is the set of all expert pairs. In this embodiment, the weight coefficient θ is between 0 and 1, which is used to control the balance between the current baseline value and the adjustment value of the model.

[0069] The consistency preference relations in the consistency preference relation matrix in this embodiment are shown in the following table:

[0070] Expert 1 Expert 2 Expert 3 Expert 4 Expert 1 0.5000 0.4048 0.5781 0.3351 Expert 2 0.6109 0.5000 0.7000 0.4156 Expert 3 0.4274 0.3400 0.5000 0.2778 Expert 4 0.7411 0.6067 0.8467 0.5000

[0071] The first column in the table above is the review expert, and the second to fifth columns are the preference information of each review expert for expert 1 to expert 4, i.e. the reviewed experts. For example, the cell in the third column of the second row indicates that expert 1's evaluation of expert 2 is 0.4048, i.e. the consistent preference relationship.

[0072] The weighted average unit 104 is used to calculate the elements corresponding to each expert in the consistent preference relationship matrix respectively to obtain the comprehensive preference information corresponding to each expert.

[0073] Among them, in the weighted average unit 104, the calculation expression of the comprehensive preference information is:

[0074]

[0075] Where p i is the comprehensive preference information of the ith expert, n is the total number of experts, and p ij is the element in the i-th row and j-th column of the consistent preference relationship matrix.

[0076] In this embodiment, it is calculated that the comprehensive preference information of expert 1 is p1=0.4545, the comprehensive preference information of expert 2 is p2=0.5567, the comprehensive preference information of expert 3 is p3=0.3863, and the comprehensive preference information of expert 4 is p4=0.6736.

[0077] The expert weight calculation unit 105 is used to normalize all comprehensive preference information to obtain the expert weight corresponding to each expert.

[0078] In this embodiment, the expert weight of expert 1 is 0.2195, the expert weight of expert 2 is 0.2688, the expert weight of expert 3 is 0.1865, and the expert weight of expert 4 is 0.3252.

[0079] The aggregation unit 106 stores an evaluation semantic set, and is used to transform the failure mode scoring matrix into a failure mode fuzzy initial evaluation matrix through the FuzzyDombi algorithm according to the expert weights and the evaluation semantic set.

[0080] In this embodiment, the evaluation semantic set is constructed according to the five-scale method, as shown in the following table:

[0081] Failure Mode Assessment Semantics Evaluation score Triangular fuzzy numbers Low Risk (VL) 1 (1,1,1) Lower risk (L) 2 (2 / 3,1,3 / 2) Medium risk (M) 3 (3 / 2,2,5 / 2) Higher risk (H) 4 (5 / 2,3,7 / 2) High risk (VH) 5 (7 / 2,4,9 / 2)

[0082] In the above table, the first column is the failure mode evaluation semantics, and the second and third columns are the evaluation scores and triangular fuzzy numbers corresponding to the failure mode evaluation semantics. For example, the cell in the second row and second column indicates that the evaluation score corresponding to low risk (VL) is 1.

[0083] Figure 3 Schematic diagram of fuzzy initial evaluation of failure modes in an embodiment of the present invention.

[0084] like Figure 3 As shown, the fuzzy initial evaluation of the failure mode calculated in this embodiment is shown in the table. The table contains the triangular fuzzy numbers of each risk factor corresponding to each failure mode. The triangular fuzzy number corresponding to each risk factor is composed of three values. These values ​​together constitute the fuzzy initial evaluation matrix of the failure mode.

[0085] The risk factor comprehensive weight calculation module 20 is used to calculate the comprehensive weight corresponding to each risk factor according to the risk factor weight evaluation matrix and the failure mode fuzzy initial evaluation matrix.

[0086] The risk factor comprehensive weight calculation module 20 includes an interval rough number matrix generation unit 201 , a subjective weight calculation unit 202 , a defuzzification unit 203 , an objective weight calculation unit 204 and a comprehensive weight calculation unit 205 .

[0087] The interval rough number matrix generating unit 201 stores an interval semantic evaluation set, and is used to convert the risk factor weight evaluation matrix into an interval rough number matrix.

[0088] The interval semantic evaluation set in this embodiment is shown in the following table:

[0089] Importance score Score meaning 1 Almost insignificant 2 Slightly affected 3，4 A lower degree of impact 5，6 More important 7 Big impact 8 Play a decisive role 9 Crucial

[0090] The first column in the table above is the importance score, and the second column is the meaning of the score corresponding to the importance score. For example, the cell in the second row and second column indicates that the meaning of the score corresponding to the importance score of 1 is almost insignificant.

[0091] The contents of the interval rough number matrix in this embodiment are shown in the following table:

[0092] expert Severity S Occurrence rate Detection degree D Repair difficulty R Repair cost C Expert 1 [8.5,9] [8,8.5] [5,6.67] [4.67,5.33] [6,7] Expert 2 [8,8.5] [5,6.67] [6,7] [6,7] [4.67,5.33] Expert 3 [8.5,9] [6,7] [8,8.5] [6,7] [4.67,5.33] Expert 4 [8.5,9] [8,8.5] [6,7] [5,6.67] [4.67,5.33]

[0093] The first column in the table above is for each expert, and the second to sixth columns are the values ​​of the scores of each expert on severity S, occurrence O, detection D, repair difficulty R, and repair cost C after transformation in the interval rough number matrix. For example, the cell in the second row and second column indicates that the score of expert 1 on severity S after transformation in the interval rough number matrix is ​​[8.5,9].

[0094] The subjective weight calculation unit 202 stores an interval rough number expectation formula, and is used to calculate the interval rough number matrix according to the interval rough number expectation formula to obtain the subjective weight corresponding to each risk factor.

[0095] In this embodiment, the subjective weights of severity S, occurrence O, detection D, repair difficulty R and repair cost C are calculated to be 0.1693, 0.2802, 0.1909, 0.1576 and 0.2020 respectively.

[0096] The defuzzification unit 203 is used to m The weighting method is used to defuzzify the failure mode fuzzy initial evaluation matrix to obtain a defuzzified evaluation matrix.

[0097] In the defuzzification unit 203, the defuzzification expression is:

[0098]

[0099] In the formula is the element in the i-th row and j-th column of the defuzzified evaluation matrix, is the element in the i-th row and j-th column of the fuzzy initial judgment matrix of the failure mode, a ij j 、a ij m and a ij u They are respectively the left value, middle value and right value of the triangular fuzzy number corresponding to the evaluation semantics of the i-th row and j-th column.

[0100] The objective weight calculation unit 204 is used to calculate the objective weight corresponding to each risk factor according to the defuzzified evaluation matrix.

[0101] In this embodiment, the percentage of the mean square value of the evaluation value of each failure mode risk factor to its standard deviation is calculated according to the defuzzified evaluation matrix to determine the degree of elimination of the difference caused by the data size, and then obtain the objective weight of the risk factor.

[0102] The percentage changes of each risk factor in this embodiment are shown in the following table:

[0103] Risk Factors Severity S Occurrence rate Detection degree D Repair difficulty R Repair cost C Coefficient of variation 0.2257 0.0315 0.2004 0.1057 0.0129

[0104] The first line in the above table shows the risk factors, and the second line shows the coefficient of variation corresponding to each risk factor, i.e., the percentage of the mean square value of the evaluation value to its standard deviation. For example, the cell in the second row and second column shows that the coefficient of variation of severity S is 0.2257. In this embodiment, the coefficient of variation is normalized to obtain the objective weight.

[0105] The comprehensive weight calculation unit 205 is used to calculate the corresponding comprehensive weight for each risk factor according to the corresponding subjective weight and objective weight.

[0106] In the comprehensive weight calculation unit 205, for each risk factor, the adjustment coefficient of the subjective weight and the adjustment coefficient of the objective weight are both set to 0.5 and combined and weighted to obtain a comprehensive weight.

[0107] The DEA evaluation score calculation module 30 stores a preset adversarial DEA model, which is used to calculate the cross efficiency value and DEA evaluation score corresponding to each failure mode through the adversarial DEA model according to the failure mode fuzzy initial evaluation matrix.

[0108] Among them, the mathematical model of the adversarial DEA model is as follows:

[0109] Objective function:

[0110]

[0111] Constraints:

[0112]

[0113] Where x id is the i-th input variable of the d-th failure mode, v id The input variable x id The associated weight, y rd is the rth output variable of the dth failure mode, u rd is the output variable y rd The relevant weights, is the d-th failure mode self-assessment efficiency value, s is the output dimension, and m is the input dimension. In this embodiment, s=2, m=2.

[0114] Among them, the input variables of the adversarial DEA model are the data corresponding to the severity S, occurrence O and detection D in the failure mode fuzzy initial evaluation matrix, and the output variables of the adversarial DEA model are the data corresponding to the repair difficulty R and repair cost C.

[0115] In this embodiment, when the failure mode obtains a smaller value in the adversarial DEA model, the failure mode is at a high priority. Considering the scale effect, the cross efficiency value x is selectedi As the output result of this method, the lower the pure technical efficiency, the more bad outputs the failure mode S, O, and D factors produce, which is a negative indicator. Therefore, the calculation expression of the DEA evaluation score is:

[0116]

[0117] In the formula is the DEA evaluation score corresponding to the ith failure mode, x i is the cross efficiency value of the ith failure mode obtained according to the adversarial DEA model. In this embodiment, the cross efficiency of the failure mode is calculated by implementing the adversarial DEA model and taking the S, O, and D factors as inputs, and the repair difficulty R and the repair cost C as the bad outputs.

[0118] The DEA evaluation scores of each failure mode in this embodiment are shown in the following table:

[0119]

[0120] In the above table, the first column is the failure mode number of each failure mode, the second column is the content of each failure mode, and the third column is the DEA evaluation score corresponding to each failure mode. For example, the cell in the second row and third column indicates that the DEA evaluation score of failure mode FM1 is 0.7579.

[0121] The comprehensive evaluation calculation module 40 is used to calculate the comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, comprehensive weight, cross efficiency value and DEA evaluation score.

[0122] The comprehensive evaluation calculation module 40 includes a support matrix generation unit 401 , an FD score calculation unit 402 , and a comprehensive evaluation calculation unit 403 .

[0123] The support matrix generating unit 401 is used to normalize the failure mode fuzzy initial evaluation matrix to obtain the failure mode fuzzy decision support matrix.

[0124] In the support matrix generation unit 401, the normalized calculation expression is:

[0125]

[0126] In the formula is the element in the i-th row and j-th column of the failure mode fuzzy decision support matrix, a ij is the element in the i-th row and j-th column of the failure mode fuzzy initial evaluation matrix.

[0127] Among them, severity S, occurrence O and detection D are positive indicators, and repair difficulty R and repair cost C in risk factors are negative indicators.

[0128] Figure 4 It is a schematic diagram of the content of the failure mode fuzzy decision support matrix in an embodiment of the present invention.

[0129] like Figure 4 As shown in the table, the first column is the failure mode, and the second to sixteenth columns are the values ​​of each risk factor corresponding to each failure mode in the failure mode fuzzy decision support matrix.

[0130] The FD score calculation unit 402 is used to aggregate the failure mode fuzzy decision support matrix through the Fuzzy Dombi method according to the comprehensive weight and the cross efficiency value, and calculate the Fuzzy Dombi evaluation value corresponding to each failure mode.

[0131] In the FD score calculation unit 402, the weight of the arithmetic mean operator and the weight of the geometric mean operator of the Fuzzy Dombi method are set to 0.8 and 0.2 respectively, and the comprehensive weight and the cross efficiency value are used as operator fusion parameters of the Fuzzy Dombi method.

[0132] The comprehensive evaluation calculation unit 403 is used to calculate the comprehensive evaluation value corresponding to each failure mode according to the Fuzzy Dombi evaluation value and the DEA evaluation score.

[0133] In the comprehensive evaluation calculation unit 403, for each failure mode, the weight of the corresponding Fuzzy Dombi evaluation value is set to 0.8, and the weight of the corresponding DEA evaluation score is set to 0.2, and the corresponding comprehensive evaluation value is obtained by weighted summation.

[0134] The fuzzy Dombi evaluation value and comprehensive evaluation value of each failure mode in this embodiment are shown in the following table:

[0135]

[0136]

[0137] In the above table, the first column is each failure mode, and the second and third columns are the Fuzzy Dombi evaluation values ​​and comprehensive evaluation values ​​corresponding to each failure mode.

[0138] The risk analysis module 50 stores an adversarial structural interpretation model and an adaptive resonance neural network, which is used to perform risk ranking and cluster analysis on all failure modes according to the comprehensive evaluation value, the Fuzzy Dombi evaluation value and the DEA evaluation score, and obtain the risk ranking result and risk rating of each failure mode as the risk analysis result.

[0139] The risk analysis module 50 includes a risk ranking unit 501 and a risk rating unit 502 .

[0140] The risk ranking unit 501 is used to rank the failure modes according to the comprehensive evaluation value. The smaller the comprehensive evaluation value is, the smaller the corresponding failure mode number is, and the greater the risk corresponding to the failure mode is.

[0141] In this embodiment, the ranking calculated by the risk ranking unit 501, i.e., the improved FMEA, is compared with the ranking obtained by the existing FMEA method, i.e., the traditional FMEA, as shown in the following table:

[0142]

[0143]

[0144] In the above table, the first column is the failure mode number, and the second and fifth columns are the content corresponding to each failure mode number, the traditional FMEA ranking, the improved FMEA ranking, and the difference between the two rankings. For example, the cell in the fourth column of the second row indicates that the risk ranking number of FM1 calculated by the risk ranking unit 501 is 6.

[0145] From the analysis of the above table, it can be seen that the failure modes with a ranking difference of greater than or equal to 10 are: FM5 main propulsion system and shafting monitoring errors, FM9 steam system monitoring errors, and FM14 imperfect fuel consumption optimization. For FM5 main propulsion system and shafting monitoring errors, the improved method, that is, this method, is considered to have a greater risk than the traditional FMEA. For the main propulsion system and shafting monitoring errors, key failure modes may be missed, thereby affecting the reliability and safety of the overall system. Improper monitoring may lead to early failures not being discovered in time, ultimately leading to system performance degradation or accidents. As for FM9 steam system monitoring errors and FM14 imperfect fuel consumption optimization, the improved method believes that the ranking of these two failure modes should be lower than the traditional method. Steam system monitoring and fuel consumption optimization are respectively located in the equipment health management module and the energy efficiency management module. Through the optimization strategy, part of the ship's loss and fixed investment can be saved. However, if the planned maintenance is not timely, the potential risk is increased, but the system operation is not affected, so its risk ranking result should be relatively backward. Therefore, the failure mode risk analysis result of the ship operation and maintenance digital twin system obtained by the failure model risk analysis device 100 is more reasonable than the existing method.

[0146] The risk rating unit 502 stores an adversarial structural explanation model and an adaptive resonance neural network, and is used to input the Fuzzy Dombi evaluation value and the DEA evaluation score corresponding to each failure mode as feature values ​​into the adversarial structural explanation model and the adaptive resonance neural network to obtain multiple clusters corresponding to different risk levels. Among them, the risk rating of the failure mode in the cluster is the risk level corresponding to the cluster.

[0147] Figure 5 is a schematic diagram of a cluster in an embodiment of the present invention.

[0148] like Figure 5 As shown in the figure, the horizontal axis is the Fuzzy Dombi index score, i.e., the Fuzzy Dombi evaluation value, and the vertical axis is the DEA index score, i.e., the DEA evaluation score. In order to adjust the clustering display effect of each failure mode, the DEA evaluation score corresponding to each failure mode is multiplied by 2 as the value of the failure mode on the vertical axis. It can be seen that a total of five clusters are generated, corresponding to five risk levels, i.e., high risk, relatively high risk, medium risk, relatively low risk, and low risk.

[0149] Among them, FM10 and FM23 are high risk, FM01, FM15 and FM18 are relatively high risk, FM02, FM03, FM04, FM05, FM06, FM9, FM11, FM12, FM13, FM16, FM17, FM19, FM20 and FM21 are medium risk, FM07, FM08 and FM14 are relatively low risk, and FM22 and FM24 are low risk.

[0150] The control module 60 stores a control program for controlling the operation of each module.

[0151] The following describes the process of using the failure model risk analysis device 100 to perform an improved FMEA failure model risk analysis method in conjunction with the accompanying drawings.

[0152] Figure 6 It is a flow chart of a failure model risk analysis method for improving FMEA in an embodiment of the present invention.

[0153] like Figure 6 As shown in the figure, the failure model risk analysis method of improving FMEA includes the following steps:

[0154] Step S1, using the initial evaluation matrix generation module 10 to calculate the failure mode fuzzy initial evaluation matrix according to the expert trust score and the failure mode score matrix.

[0155] Step S2: The risk factor comprehensive weight calculation module 20 is used to calculate the comprehensive weight corresponding to each risk factor according to the risk factor weight evaluation matrix and the failure mode fuzzy initial evaluation matrix.

[0156] Step S3, using the DEA evaluation score calculation module 30 to calculate the cross efficiency value and DEA evaluation score corresponding to each failure mode through the adversarial DEA model according to the failure mode fuzzy initial evaluation matrix.

[0157] Step S4, using the comprehensive evaluation calculation module 40 to calculate the fuzzy Dombi evaluation value and the comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, comprehensive weight, cross efficiency value and DEA evaluation score.

[0158] Step S5, using the risk analysis module 50 to perform risk ranking and cluster analysis on all failure modes based on the comprehensive evaluation value, DEA evaluation score and Fuzzy Dombi evaluation value, combined with the adversarial structure interpretation model and the adaptive resonance neural network, to obtain the risk ranking result and risk rating of each failure mode as the risk analysis result.

[0159] Functions and Effects of the Embodiments

[0160] According to the improved FMEA failure model risk analysis device and method involved in this embodiment, through the initial evaluation matrix generation module, the expert weight is calculated based on the preference relationship, so as to fully consider the difference in the expert's preference strength, reduce the influence of this factor on the expert's decision, and make the decision more real and reliable; through the risk factor comprehensive weight calculation module, the interval rough number is used to obtain the subjective weight according to the risk factor weight evaluation matrix, and the logarithmic percentage change driven weighting method is used to obtain the objective weight, and then the comprehensive weight is obtained; through the comprehensive evaluation calculation module and the risk analysis module, the adversarial data envelopment analysis and the Fuzzy Dombi method are used to sort the failure modes, and the advantages and disadvantages of the failure modes are evaluated more comprehensively, and the results are more reliable, and the adaptive resonance neural network is used for cluster analysis. In short, the device and method can generate more reliable and accurate failure mode risk analysis results.

[0161] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An improved FMEA failure model risk analysis device, used for obtaining risk analysis results of all failure modes according to a failure mode scoring matrix formed by scores of each expert on each risk factor of each failure mode, a risk factor weight evaluation matrix formed by scores of each expert on each risk factor, and expert trust scores of all experts, characterized in that: include: An initial evaluation matrix generation module is used to calculate a failure mode fuzzy initial evaluation matrix according to the expert trust score and the failure mode score matrix; A risk factor comprehensive weight calculation module, used to calculate the comprehensive weight corresponding to each of the risk factors according to the risk factor weight evaluation matrix and the failure mode fuzzy initial evaluation matrix; A DEA evaluation score calculation module stores a preset adversarial DEA model, and is used to calculate the cross efficiency value and DEA evaluation score corresponding to each failure mode through the adversarial DEA model according to the failure mode fuzzy initial evaluation matrix; A comprehensive evaluation calculation module, used for calculating the fuzzy Dombi evaluation value and the comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, the comprehensive weight, the cross efficiency value and the DEA evaluation score; The risk analysis module stores an adversarial structural interpretation model and an adaptive resonance neural network, and is used to perform risk sorting and cluster analysis on all the failure modes according to the comprehensive evaluation value, the DEA evaluation score and the Fuzzy Dombi evaluation value, so as to obtain the risk sorting result and risk rating of each failure mode as the risk analysis result.

2. The improved FMEA failure model risk analysis device according to claim 1, Features: Wherein, the initial evaluation matrix generation module includes: A preference relationship matrix generating unit stores a preset correction formula, and is used to calculate the trust scores of all the experts according to the correction formula to obtain a preference relationship matrix containing the preference information corresponding to each other among the experts; A feature preference relationship matrix generating unit, used for calculating a feature preference relationship matrix according to the preference relationship matrix; A consistency judgment unit, storing preset judgment conditions and adjustment formulas, for judging whether the feature preference relationship matrix satisfies the preset judgment conditions, and if so, taking the feature preference relationship matrix as a consistent preference relationship matrix; if not, adjusting the feature preference relationship matrix according to the adjustment formula, and re-executing the consistency judgment unit; A weighted average unit, used to calculate the elements corresponding to each of the experts in the consistent preference relationship matrix respectively, to obtain the comprehensive preference information corresponding to each of the experts; An expert weight calculation unit, used for normalizing all the comprehensive preference information to obtain the expert weight corresponding to each of the experts; An aggregation unit stores an evaluation semantic set and is used to convert the failure mode scoring matrix into the failure mode fuzzy initial evaluation matrix through a FuzzyDombi algorithm according to the expert weight and the evaluation semantic set.

3. The improved FMEA failure model risk analysis device according to claim 2, characterized in that: in, In the feature preference relationship matrix generating unit, the preference relationship matrix middle The calculation expression is: In the formula is the preference relationship matrix The element in row i and column j in p i+s,i+s+1 is the preference relationship matrix P = (p ij ) n×n The element at the i+sth row and i+s+1th column in In the consistency judgment unit, the expression of the preset judgment condition is: CI(P (t) )≤δ0, Where P (t) is the current feature preference relationship matrix, CI(P (t) ) is the feature preference relationship matrix P (t) The consistency index, δ0 ​​is the consistency judgment threshold, The expression of the adjustment formula is: In the formula is the element in the i-th row and j-th column of the adjusted feature preference relationship matrix, is the element in the i-th row and j-th column of the preference relationship matrix before adjustment, is the element in the i-th row and j-th column of the feature preference relationship matrix before adjustment, θ is the weight coefficient, N is the set of all expert pairs, In the weighted average unit, the calculation expression of the comprehensive preference information is: Where p i is the comprehensive preference information of the ith expert, n is the total number of experts, and p ij is the element in the i-th row and j-th column of the consistent preference relationship matrix.

4. The improved FMEA failure model risk analysis device according to claim 1, Features: Wherein, the risk factor comprehensive weight calculation module includes: An interval rough number matrix generating unit storing an interval semantic evaluation set and used for converting the risk factor weight evaluation matrix into an interval rough number matrix; A subjective weight calculation unit stores an interval rough number expectation formula, and is used to calculate the interval rough number matrix according to the interval rough number expectation formula to obtain the subjective weight corresponding to each risk factor; Defuzzification unit is used to m Defuzzifying the failure mode fuzzy initial evaluation matrix by a weighting method to obtain a defuzzified evaluation matrix; An objective weight calculation unit, used for calculating the objective weight corresponding to each of the risk factors according to the defuzzified evaluation matrix; The comprehensive weight calculation unit is used to calculate the corresponding comprehensive weight for each risk factor according to the corresponding subjective weight and the objective weight.

5. The improved FMEA failure model risk analysis device according to claim 4, characterized in that: in, In the defuzzification unit, the defuzzification expression is: In the formula is the element in the i-th row and j-th column of the defuzzified evaluation matrix, is the element in the i-th row and j-th column of the fuzzy initial judgment matrix of the failure mode, a ij l 、a ij m and a ij u They are the left value, middle value and right value of the triangular fuzzy number corresponding to the evaluation semantics of the i-th row and j-th column, For each of the risk factors, the adjustment coefficient of the subjective weight and the adjustment coefficient of the objective weight are both set to 0.5 and combined and weighted to obtain the comprehensive weight.

6. The improved FMEA failure model risk analysis device according to claim 1, characterized in that: in, The mathematical model of the adversarial DEA model is as follows: Objective function: Constraints: Where x id is the i-th input variable of the d-th failure mode, v id The input variable x id The associated weight, y rd is the rth output variable of the dth failure mode, u rd is the output variable y rd The relevant weights, is the self-evaluation efficiency value of the d-th failure mode, s is the output dimension, m is the input dimension, The risk factors include severity S, occurrence O, detection D, repair difficulty R and repair cost C. The input variables are severity S, occurrence O and detection D. The output variables are the repair difficulty R and the repair cost C, The calculation expression of the DEA evaluation score is: In the formula is the DEA evaluation score corresponding to the ith failure mode, x i is the cross efficiency value obtained for the ith failure mode according to the adversarial DEA model.

7. The improved FMEA failure model risk analysis device according to claim 1, Features: Wherein, the comprehensive evaluation calculation module includes: A support matrix generation unit, used for normalizing the failure mode fuzzy initial evaluation matrix to obtain a failure mode fuzzy decision support matrix; An FD score calculation unit, configured to aggregate the failure mode fuzzy decision support matrix by a fuzzy Dombi method according to the comprehensive weight and the cross efficiency value, and calculate the fuzzy Dombi evaluation value corresponding to each failure mode; A comprehensive evaluation calculation unit is used to calculate the comprehensive evaluation value corresponding to each failure mode according to the Fuzzy Dombi evaluation value and the DEA evaluation score.

8. The improved FMEA failure model risk analysis device according to claim 7, characterized in that: in, In the support matrix generation unit, the normalized calculation expression is: In the formula is the element in the i-th row and j-th column of the failure mode fuzzy decision support matrix, a ij is the element in the i-th row and j-th column of the failure mode fuzzy initial evaluation matrix, The severity S, occurrence O and detection D of the risk factors are the positive indicators. The repair difficulty R and the repair cost C in the risk factors are the negative indicators. In the FD score calculation unit, the weight of the arithmetic mean operator and the weight of the geometric mean operator of the Fuzzy Dombi method are set to 0.8 and 0.2 respectively. The comprehensive weight and the cross efficiency value are used as operator fusion parameters of the Fuzzy Dombi method. In the comprehensive evaluation calculation unit, for each failure mode, the weight of the corresponding Fuzzy Dombi evaluation value is set to 0.8, and the weight of the corresponding DEA evaluation score is set to 0.2, and the corresponding comprehensive evaluation value is obtained by weighted summation.

9. The improved FMEA failure model risk analysis device according to claim 1, Features: Wherein, the risk analysis module includes: A risk ranking unit, used for ranking the failure modes according to the comprehensive evaluation value; a risk rating unit storing the adversarial structural explanation model and the adaptive resonance neural network, and used to input the Fuzzy Dombi evaluation value and the DEA evaluation score corresponding to each of the failure modes as feature values ​​into the adversarial structural explanation model and the adaptive resonance neural network to obtain a plurality of clusters corresponding to different risk levels; The smaller the comprehensive evaluation value is, the smaller the corresponding failure mode number is, and the greater the risk corresponding to the failure mode is. The risk rating of the failure mode in the cluster is the risk level corresponding to the cluster.

10. An improved FMEA failure model risk analysis method is used to obtain risk analysis results of all failure modes based on a failure mode scoring matrix formed by scores of each expert on each risk factor of each failure mode, a risk factor weight evaluation matrix formed by scores of each expert on each risk factor, and expert trust scores of all experts, characterized in that: include: Step S1, calculating a failure mode fuzzy initial evaluation matrix according to the expert trust score and the failure mode score matrix; Step S2, calculating the comprehensive weights corresponding to the risk factors according to the risk factor weight evaluation matrix and the failure mode fuzzy initial evaluation matrix; Step S3, according to the failure mode fuzzy initial evaluation matrix, the cross efficiency value and DEA evaluation score corresponding to each failure mode are calculated by the adversarial DEA model; Step S4, calculating the fuzzy Dombi evaluation value and the comprehensive evaluation value corresponding to each failure mode according to the failure mode fuzzy initial evaluation matrix, the comprehensive weight, the cross efficiency value and the DEA evaluation score; Step S5, based on the comprehensive evaluation value, the DEA evaluation score and the Fuzzy Dombi evaluation value, combined with the adversarial structure interpretation model and the adaptive resonance neural network, risk ranking and cluster analysis are performed on all the failure modes to obtain the risk ranking result and risk rating of each failure mode as the risk analysis result.

Images