Aeroengine fault diagnosis method and system based on extension association rule mining
By adopting the extended correlation rule mining method in aircraft engine fault diagnosis, combining Pearson correlation coefficient and extended transformation algorithm, the problems of low efficiency and invalid rules in the existing technology are solved, and multi-dimensional evaluation and efficient fault detection are realized.
Patent Information
- Application Number
- CN202111063095.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-10
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-09-10
AI Technical Summary
The existing correlation rule mining algorithm is inefficient in aircraft engine fault diagnosis, and it is impossible to effectively screen out inference rules, and the evaluation indicators are single, making it easy to generate invalid rules.
The method of mining based on the extension association rules is used to calculate the fault feature correlation through Pearson correlation coefficient, and FP-Tree is generated by combining frequent item set mining and extension transformation algorithm. The association rules are sorted using the extension superiority evaluation method to comprehensively analyze the faults. Multi-dimensional indicators such as frequency, flight safety, and economic applicability.
It improves the efficiency of aircraft engine fault detection, broadens the evaluation indicators for mining related rules, improves the effectiveness and practicality of fault analysis, and reduces the occurrence of invalid rules.
Smart Images

Figure CN113761032B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of aircraft engine fault diagnosis, and in particular relates to an aircraft engine fault diagnosis method and system based on extension association rule mining. Background Art
[0002] The aircraft engine is the heart of the aircraft, providing power for flight missions. Its reliability and safety are of vital importance. Due to the complex structure of aircraft engines and the different service life of components, the causes of failures are diverse, which increases the difficulty of aircraft engine fault diagnosis. However, a large amount of operation and maintenance data has been accumulated in the actual process. The implicit correlation can be analyzed through machine learning models, thereby providing effective support for aircraft engine fault diagnosis. Operation and maintenance data comes from a wide range of sources, has different structures, and has low value density. It is necessary to use association rule mining technology to extract strong association rules from it for aircraft engine fault association analysis.
[0003] The classic association rule mining algorithm is the Apriori algorithm, which prioritizes the search in the width direction and generates frequent item sets by traversing layer by layer. However, the data set required to be traversed is too large, resulting in low operating efficiency. Therefore, the FP-Tree algorithm was proposed. This algorithm prioritizes the search in the depth direction and generates frequent items by inserting tree nodes, which has a faster operating efficiency. However, the FP-Tree algorithm also has certain defects: the algorithm process needs to traverse the transaction set multiple times, and the operating efficiency still needs to be improved; association rule mining is prone to invalid rules, and constraints need to be added for the fault diagnosis field; the main evaluation indicators of association rule mining based on FP-Tree are confidence and support, which overemphasizes frequency characteristics and has low utility value in the fault diagnosis field. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that the existing aircraft engine fault detection efficiency based on association rule mining algorithm is low and effective inference rules cannot be screened out in terms of fault diagnosis. An aircraft engine fault diagnosis method and system based on extension association rule mining are proposed.
[0005] An aircraft engine fault diagnosis method based on extension association rule mining, comprising:
[0006] Step 1: Obtain historical data during the operation of the aircraft engine, the historical data including life calculation, flight records, faulty parts status and supply chain information; establish an aircraft engine fault transaction set B based on the historical data, that is, B = {B1, B2, ..., Bn}, where B1, B2, ..., Bn n Indicates different fault transaction types, and expands different fault transaction types into different fault features, namely B i = {Oi ,{(C1,V1),(C2,V2),...,(C n ,V n )}}, where O i Indicates the name of the i-th faulty component; C i is the feature of the i-th faulty component, V i Represents C i The characteristic value of i = 1, 2, ..., n, where n represents the number of faulty component characteristics;
[0007] Step 2: Use the Pearson correlation coefficient to calculate the correlation between every two fault features, set a correlation threshold, remove one fault feature in the fault feature pair that is higher than the correlation threshold, and use the remaining fault features as key fault features;
[0008] Step 3: Use the frequent item set mining algorithm to extract frequent items from key fault features, and extract strong association rules from frequent items. The specific process includes:
[0009] Step 31: Set the minimum support threshold Sup_min, and combine the fault transactions that are greater than or equal to the minimum support threshold into a frequent 1-item set;
[0010] Step 32: Sort the fault features in the frequent 1-item set in descending order of support to form an ordered set B s ;
[0011] Step 3. Use the extension transformation algorithm to eliminate the infrequent transaction set B0 from the aircraft engine failure transaction set to obtain the set B-B0;
[0012] Step 34: According to the ordered set B s Sort each transaction in the set B-B0 in the order of , and generate an ordered frequent item set [p|P], where p is the first item in the set [p|P] and P is the remaining item;
[0013] Step 35: Take T as the root node and use the recursive method to embed the node ([p|P], T) to generate FP-Tree;
[0014] Step 36: Generate conditional pattern bases based on the nodes of FP-Tree, and mine all frequent item sets from the conditional pattern bases;
[0015] Step 37: Calculate the confidence of all frequent items in the frequent item set, set the confidence threshold, take the frequent items whose confidence is higher than the confidence threshold, and extract strong association rules based on the frequent items higher than the confidence threshold;
[0016] Step 4: Sort the strong association rules based on the extension superiority evaluation method, and diagnose the faults according to the sequence of the strong association rules.
[0017] An aircraft engine fault diagnosis system based on extension association rule mining comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, an aircraft engine fault diagnosis method based on extension association rule mining is implemented.
[0018] The beneficial effects of the present invention are:
[0019] The present invention can solve the problems of single evaluation index, easy generation of invalid rules, low mining efficiency, etc. of traditional association rule mining. The present invention integrates the extension transformation method, extension goodness evaluation method and association rule mining algorithm, conducts comprehensive analysis from multiple dimensions such as fault frequency characteristics, flight safety, economic applicability, maintainability, etc., and effectively sorts the association rules through extension goodness, which greatly improves the efficiency of aircraft engine fault detection, broadens the evaluation index of association rule mining, and improves the effectiveness and practicality of aircraft engine fault analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a flow chart of the method of the present invention;
[0021] Figure 2 Modeling a flow chart for an aircraft engine failure transaction set;
[0022] Figure 3 Diverge ER relationships for extensible transactions;
[0023] Figure 4 It is the process of constructing FP-Tree for extension transformation;
[0024] Figure 5 It is composed of the structure of transaction set B;
[0025] Figure 6 This is an extension goodness evaluation process for engine failure risk assessment. DETAILED DESCRIPTION
[0026] Specific implementation method one: Extenics is an original cross-disciplinary subject, in which the extenic primitive model objectively describes data characteristics from the perspectives of objects, features, and quantities, providing strong support for transaction set modeling; the extenic transformation method can effectively transform the domain, simplify the traversal space of FP-Tree, and improve operating efficiency; the extenic superiority evaluation method is mainly used for the expansion of multi-level and multi-dimensional evaluation indicators, providing a new idea for the utility evaluation of association rule mining.
[0027] To this end, the present invention proposes an aircraft engine fault analysis method based on extension association rule mining, introduces theories such as extension primitive model, extension transformation method, and extension goodness evaluation method, improves the mining efficiency of FP-Tree association rule mining algorithm, and sorts the priority of association rules from multi-dimensional evaluation indicators, improves the utility value of reasoning results, and better guides the quality tracing of fault diagnosis.
[0028] Combination Figure 1-6 Specifically describing this embodiment, this embodiment is an aircraft engine fault analysis method based on extension association rule mining, comprising the following steps:
[0029] Step 1: Obtain historical data of the aircraft engine operation and maintenance process, wherein the historical data includes life calculation, flight records, faulty component status, and supply chain information; based on the historical data of the aircraft engine operation and maintenance process, establish an aircraft engine fault transaction set B, where B = (O, C, V), where O represents the name of the faulty component; C represents the faulty component feature, characterizing the fault phenomenon; and V represents the feature value, which is used for quantitative analysis of faulty transactions; diverge processing is performed on the aircraft engine fault transaction set according to the multi-object multi-feature divergence rule; the specific process includes:
[0030] like Figure 3 As shown in the figure, combined with the configuration of the aircraft engine, the aircraft engine fault transaction set B can be expanded into different fault types by using the multi-object divergence method, that is, B = {B1, B2, ..., Bn}, where B1, B2, ..., Bn n Indicates different fault types, for example, high-pressure compressor fault B1, low-pressure compressor fault B2, high-pressure turbine fault B3, low-pressure turbine fault B4, inlet fault B5 and accessory fault B6, etc.; then according to the neutral material list, engine maintenance record and fault information record, the multi-feature divergence method can be used to expand the fault transaction into different fault features, namely B i = {O i ,{(C1,V1),(C2,V2),...,(C n ,V n )}}, where O i Indicates the name of the i-th faulty component; O i ,{(C1,V1),(C2,V2),...,(C n ,V n )} represents the fault transaction details of the i-th faulty component, C i is the feature of the i-th faulty component, V i Represents C iThe characteristic value of i = 1, 2, ..., n, where n represents the number of faulty component characteristics; for example, the number of impacts C1, the consequence of the fault C2, the engine type C3, the location of the faulty component C4, the batch of the faulty component C5, the working time of the faulty component C6, the nature of the fault C7, etc.;
[0031] Step 2: Use the Pearson correlation coefficient to calculate the linear correlation between every two fault features, set a correlation threshold, and use the fault features above the correlation threshold as key fault features;
[0032] Aircraft engine failure transactions contain a large amount of multi-source heterogeneous data. The same feature information may be expressed in different ways, such as failure consequences and failure nature. In order to solve the problem of feature expression consistency, the correlation analysis method is used to prune different fault features, that is, the Pearson correlation coefficient (Pearson) is used to calculate the linear correlation between fault features. Fault features with a Pearson coefficient higher than the minimum threshold are highly linearly correlated. It is necessary to reduce the dimension of the number of features for subsequent association rule mining.
[0033] Through the multi-object multi-feature divergence method, the aircraft engine failure events are described in detail from qualitative and quantitative perspectives, and the multi-source heterogeneous original failure data are effectively integrated;
[0034] Step 3: Perform frequent item set mining on key fault features, analyze the high-frequency relationship between aircraft engine fault features, regard each fault feature as an item, and a historical record containing multiple fault features is a fault transaction. Multiple fault transactions constitute an aircraft engine fault primitive. Through frequent item set mining, the implicit relationship between fault transactions can be discovered; first, establish a mapping relationship between aircraft engine fault transactions and frequent pattern trees (FP-Tree), then integrate the extension transformation method into the traditional FP-Tree structure, which can optimize the frequent item search process, thereby improving the algorithm mining efficiency, and finally, set the confidence threshold, and filter out strong association rules from frequent items according to the confidence; Among them, the FP-Tree algorithm does not need to traverse layer by layer along the width direction, but generates tree nodes along the depth direction to store frequent item information. The algorithm has good mining efficiency when facing large-scale fault data sets;
[0035] In step 3, frequent item sets are mined for key fault features. The specific process includes:
[0036] Step 31: Set the minimum support threshold Sup_min and scan the aircraft engine failure transaction set B; the support represents the frequency of occurrence of transaction combinations. The higher the support, the more likely the transaction is to occur. Filter out high-frequency transactions according to the minimum support;
[0037] Step 3.2: Take the fault features higher than Sup_min as frequent 1-item sets and arrange them in descending order of support to generate an ordered set B. s ;
[0038] Step 33: Optimize the search space of the aircraft engine aircraft engine failure transaction set B, that is, use the extension transformation method to eliminate the infrequent transaction set B0 to obtain the set B-B0;
[0039] Step 34: Scan the optimized transaction set B-B0 again and follow the set B s Sort the included items of each transaction in the order of , and generate the set [p|P], where p is the first item in the set [p|P] and P is the remaining item;
[0040] Step 35: Take T as the root node and construct FP-Tree by recursively calling child nodes (recursive calling is a commonly known method in computer programming), that is, embedding node ([p|P], T);
[0041] Step 36: Generate a conditional pattern base based on the node path of FP-Tree, and mine all frequent item sets from the conditional pattern base;
[0042] Step 37: Calculate the confidence of all frequent items in the frequent item set, set the confidence threshold, take the frequent items with confidence higher than the confidence threshold, and extract strong association rules based on the frequent items higher than the confidence threshold; the association rule represents the probability of launching another frequent item set under the condition of a certain frequent item set. If the confidence of the association rule is greater than or equal to the minimum confidence, it is a strong association rule);
[0043] From the above process, it can be seen that the fault association rule generation based on the extension transformation FP-Tree uses the extension transformation method to eliminate the invalid search space of the original transaction set, thereby improving the construction efficiency of the FP-Tree;
[0044] Step 4: Sort the strong association rules based on the extension superiority evaluation method, and diagnose the faults according to the sequence of the strong association rules.
[0045] Specific implementation method 2: This implementation method is different from the specific implementation method 1 in that the linear correlation between every two fault features is calculated using the Pearson correlation coefficient (Pearson), and the calculation formula is as follows:
[0046]
[0047] in, Represents feature C X and feature C Y relevance; Represents feature C XThe standard deviation of Represents feature C Y The standard deviation of ; cov represents covariance; X and Y are from 1 to n, X≠Y.
[0048] Through the Pearson correlation analysis method, pruning is performed to eliminate the fault features with high linear correlation and reduce invalid feature items.
[0049] The other steps and parameters are the same as those in the first embodiment.
[0050] Specific implementation method three: This implementation method is different from specific implementation methods one or two in that the step three uses an extension transformation method to eliminate the infrequent transaction set B0 to obtain a set B-B0. The specific process includes:
[0051] When scanning the fault transaction set B, the traditional FP-Tree will traverse all transactions one by one, which consumes a lot of time. Therefore, the present invention adds a detachable transformation algorithm based on the traditional FP-Tree structure.
[0052] Assume that there is a subset B0 in the transaction set B, which can be expressed as:
[0053] B0={T 01 ,T 02 …T 0m}
[0054]
[0055] Among them, T 01 ,T 02 …T 0m It is defined as a non-frequent transaction, that is, all the features contained in the transaction are non-frequent items. From the FP-Tree algorithm steps, it can be seen that the frequent item set does not contain non-frequent items, that is, there is no need to scan non-frequent transactions T 01 ,T 02 …T 0m ;
[0056] Therefore, the extension transformation method is used to optimize the original search space of the transaction set B, and the subset B0 composed of infrequent transactions is eliminated, which can be expressed as:
[0057] T - B=B-B0
[0058] Among them, T - It represents the reduction transformation in the extension transformation, which indicates that the traversal range of the original transaction set is reduced during the domain transformation process, thereby improving the construction efficiency of the FP-Tree.
[0059] The other steps and parameters are the same as those in the first or second embodiment.
[0060] Specific implementation method 4: This implementation method is different from any one of specific implementation methods 1 to 3 in that step 4 sorts the strong association rules based on the extension superiority evaluation method, and processes the faults according to the priority order according to the sorted strong association rules. The specific process includes:
[0061] Through the extension transformation FP-Tree algorithm, frequent items of faulty parts are mined, thereby mining strong association rules between faulty transactions; strong association rules refer to rules with high conditional probability, which represent that the reasoning results have high confidence. However, the occurrence of engine failures is often strongly associated with multiple fault features. Multiple rules can be inferred from the same constraint antecedent. For example, a certain engine type and faulty part location failure may be related to multiple batches of faulty parts. For multiple strong association rules, the traditional method is to use confidence, support and other indicators for priority sorting. However, for the operation and maintenance process of aviation engines, it is inappropriate to rely solely on the fault frequency characteristics for maintenance guidance. The maintainability, flight safety, and economic applicability of the faulty parts are also required. Therefore, the strong association rules mined based on the extension transformation FP-Tree still need to be further analyzed, and multiple indicators are used to comprehensively sort the maintenance priority (the priority of the strong association rules) to screen out the strong association rules that meet the maintenance needs.
[0062] The extension superiority evaluation method expands the evaluation index based on the support and confidence analysis by constructing the extension distance and correlation function, which provides an effective way for the multi-dimensional evaluation of the reasoning results. For different evaluation indicators, the extension superiority evaluation method is used to establish the standard correlation degree, and the superiority value is used to characterize the risk degree of engine failure in different dimensions, and further sort the fault priority.
[0063] The extension superiority evaluation method selects diversified measurement indicators, focusing on the feasibility and stability of the indicators. The evaluation process includes the selection of measurement indicators, weight coefficient allocation, initial evaluation, correlation and superiority calculation, etc.
[0064] Step 41: Construction of engine failure risk assessment indicator set:
[0065] Aircraft engine failures involve multiple links such as aircraft management, engine management, field maintenance, and factory overhaul. Combined with the engine configuration and maintenance manual, we reasonably select evaluation indicators closely related to engine failures and establish an engine failure risk evaluation indicator set, which can be expressed as:
[0066]
[0067] The engine failure risk assessment index set MI includes n 1 MI sub-indicators i , using the characteristic element MI i ={Ci ,V i} Characterization; For the risk assessment of aircraft engine failure, it specifically includes frequency characteristics, flight safety, economic applicability, and maintainability. Among them, the frequency characteristics can be divided into support and confidence, the economic applicability can be divided into parts cost and maintenance hours, and the maintainability can be divided into the latter constraint and the number of replacement parts;
[0068] Step 4.2: Assign weights to the engine failure risk assessment indicators and set the assessment indicators α that must be met. i ;
[0069] Step 43: Based on the evaluation index α that must be met i , remove those below α i Association rules of
[0070] Step 44: Eliminate the samples below α according to step 43 i The association rules are constructed, the corresponding extension association function is constructed, the multi-index extension association degree is obtained based on the extension association function, and the multi-index extension association degree is normalized;
[0071] Step 4 and 5: Based on the normalized multi-index extensible correlation, the fault risk items are obtained and sorted according to their merits. The item with the largest merit indicates that this fault needs to be handled with priority.
[0072] The other steps and parameters are the same as those in Specific Embodiments 1 to 3.
[0073] Specific implementation method 5: This implementation method is different from specific implementation methods 1 to 4 in that the step 42 performs weight allocation on the engine failure risk evaluation index and sets the evaluation index α that must be satisfied. i , specifically:
[0074] Multiple evaluation index MI i Failure risk is evaluated from multiple dimensions, but different indicators MI i There are differences in the evaluation system. For example, for a certain failure of an aircraft engine, the importance of flight safety is often higher than economic applicability. Therefore, there are differences in the weights of evaluation indicators. The weight distribution of multiple evaluation indicators can be expressed as:
[0075]
[0076] Among them, α is the weight coefficient set, corresponding to the weight of each sub-indicator of the evaluation index set MI; k represents the weight ordinal number; Λ Indicates the evaluation index that must be met; α i are the evaluation indicators that must be met, except α i The sum of the weight coefficients of other evaluation indicators is 1.
[0077] The other steps and parameters are the same as those in Specific Embodiments 1 to 4.
[0078] Specific implementation method 6: This implementation method is different from the specific implementation methods 1 to 5 in that the step 43 is based on the evaluation index α that must be satisfied. i , remove those below α i The specific process of association rules is as follows:
[0079] From a statistical point of view, confidence is the calculation of the conditional probability of the inference result. The association rules with confidence lower than the threshold indicate that there is no correlation between the preceding and following items of the fault characteristics and should be eliminated. Therefore, the first evaluation of fault risk must meet the evaluation index α. i , the calculation method of the first evaluation index can be expressed as follows:
[0080]
[0081] in, Association rules The correlation between the reasoning result and the fault phenomenon is an indicator that must be met.
[0082] The other steps and parameters are the same as those in Specific Implementations 1 to 5.
[0083] Specific implementation method 7: This implementation method is different from the specific implementation methods 1 to 6 in that the step 44 removes the cells below α in step 43. i The association rules are constructed, the corresponding extension association function is constructed, the multi-index extension association degree is obtained based on the extension association function, and the multi-index extension association degree is normalized. The specific process includes:
[0084] Assume that the failure risk item set Z = {Z1, Z2…Z m} all meet the first evaluation, then the multi-index extension correlation can be expressed as follows:
[0085] K i =(K i (Z1),K i (Z2)…K i (Z m )),i=1,2,…n
[0086] Where m represents the total number of failure risk items;
[0087] Normalize the multi-index extension correlation:
[0088]
[0089] Among them, k ij represents the normalized extension correlation;
[0090] The standard correlation of each evaluation index of the fault risk item can be expressed as:
[0091] k i =(k i1 ,k i2 …k im ),i=1,2…n
[0092] The other steps and parameters are the same as those in the specific implementation modes 1 to 6.
[0093] Specific implementation eight: This implementation differs from specific implementations one to seven in that in steps four and five, fault risk items are obtained based on the multi-index extension correlation after normalization, and are sorted according to the merits of the fault risk items. The specific process includes:
[0094] For the failure risk item set Z = {Z1, Z2…Z m}, using the evaluation index MI = {MI1, MI2…MI n}Evaluate each fault risk item in multiple dimensions, and the standard correlation matrix can be expressed as:
[0095]
[0096] Design the fault risk item Z according to the aircraft engine fault operation and maintenance requirements j The goodness calculation formula is as follows:
[0097]
[0098] Among them, P1(Z j ) represents the weighted merit of each evaluation index, P2(Z j ) indicates that the goodness of each evaluation index takes the minimum value, P3(Z j ) indicates that the goodness of each evaluation index takes the maximum value.
[0099] For the failure risk item set Z = {Z1, Z2…Z m The goodness P(Z j ) and sorted by the degree of excellence. The item with the largest excellence indicates that the fault needs to be handled first, as shown below:
[0100]
[0101] The extension goodness evaluation method for engine failure risk assessment conducts a comprehensive analysis from the dimensions of failure frequency characteristics, flight safety, economic applicability, maintainability, etc., broadens the evaluation indicators of association rule mining, sorts the association rules by extension goodness, and represents the processing priority of different fault transactions, which is used to assist the maintenance decision of aircraft engines.
[0102] The other steps and parameters are the same as those in Specific Embodiments 1 to 7.
[0103] Specific embodiment nine: This embodiment is an aircraft engine fault diagnosis system based on extension association rule mining, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that when the processor executes the computer program, the steps of any method described in specific embodiments one to eight are implemented.
Claims
1. An aircraft engine fault diagnosis method based on extension association rule mining, characterized in that: include: Step 1: Obtain historical data during the operation of the aircraft engine, the historical data including life calculation, flight records, faulty parts status and supply chain information; establish an aircraft engine fault transaction set B based on the historical data, that is, B = {B1, B2, ..., Bn}, where B1, B2, ..., Bn n Indicates different fault transaction types, and expands different fault transaction types into different fault features, namely B i = {O i ,{(C1,V1),(C2,V2),...,(C n ,V n )}}, where O i Indicates the name of the i-th faulty component; C i is the feature of the i-th faulty component, V i Represents C i The characteristic value of i = 1, 2, ..., n, where n represents the number of faulty component characteristics; Step 2: Use the Pearson correlation coefficient to calculate the correlation between every two fault features, set a correlation threshold, remove one fault feature in the fault feature pair that is higher than the correlation threshold, and use the remaining fault features as key faults; Step 3: Use the frequent item set mining algorithm to extract frequent items from key fault features, and extract strong association rules from frequent items. The specific process includes: Step 31: Set the minimum support threshold Sup_min, and combine the fault transactions that are greater than or equal to the minimum support threshold into a frequent 1-item set; Step 32: Sort the fault features in the frequent 1-item set in descending order of support to form an ordered set B s ; Step 3. Use the extension transformation algorithm to eliminate the infrequent transaction set B0 from the aircraft engine failure transaction set to obtain the set B-B0; Step 34: According to the ordered set B s Sort each transaction in the set B-B0 in the order of , and generate an ordered frequent item set [p|P], where p is the first item in the set [p|P] and P is the remaining item; Step 35: Take T as the root node and use the recursive method to embed the node ([p|P], T), and the node generates FP-Tree; Step 36: Generate conditional pattern bases based on the nodes of FP-Tree, and mine all frequent item sets from the conditional pattern bases; Step 37: Calculate the confidence of all frequent items in the frequent item set, set the confidence threshold, take the frequent items whose confidence is higher than the confidence threshold, and extract strong association rules based on the frequent items higher than the confidence threshold; Step 4: Sort the strong association rules based on the extension superiority evaluation method, and diagnose the faults according to the sequence of the strong association rules.
2. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 1 is characterized in that: In step 2, the Pearson correlation coefficient is used to calculate the correlation between every two fault features, and the formula is as follows: in, Represents feature C X and feature C Y relevance; Represents feature C X The standard deviation of Represents feature C Y The standard deviation of ; cov represents covariance; X and Y are from 1 to n, X≠Y.
3. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 2 is characterized in that: The step 33 uses the extension transformation method to eliminate the infrequent transaction set B0 from the aircraft engine failure transaction set to obtain the set B-B0. The specific process includes: Assuming that there is an infrequent transaction set B0 in the fault transaction set B, the infrequent transaction set B0 is expressed as: B0={T 01 ,T 02 …T 0m } Among them, T 01 ,T 02 …T 0m For infrequent transactions; Eliminating the subset B0 consisting of infrequent transactions, it is expressed as: T - B=B-B0 Among them, T - Represents the subtraction transformation in extension transformation.
4. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 3 is characterized in that: The step 4 is to sort the strong association rules based on the extension superiority evaluation method, and the specific process includes: Step 4.1: Construct an engine failure risk assessment index set, denoted as in, is the engine failure risk assessment index set n 1 Each sub-indicator corresponds to a fault transaction. i ={C i ,V i }; Step 4.2: Assign weights to the engine failure risk assessment indicators and set the assessment indicators α that must be met. i ; Step 43: Based on the evaluation index α that must be met i , remove those below α i The association rules are kept as follows: i Association rules of Step 4. For greater than or equal to α i The association rules are constructed, the corresponding extension association function is constructed, the multi-index extension association degree is obtained based on the extension association function, and the multi-index extension association degree is normalized; Step 4 and 5: Based on the normalized multi-index extensible correlation, the fault risk items are obtained and sorted according to their merits. The item with the largest merit is the fault that needs to be handled first.
5. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 4 is characterized in that: The step 42 performs weight distribution on the engine failure risk evaluation index, and sets the evaluation index α that must be met. i , specifically: Among them, α is the weight coefficient set, corresponding to the weight of each sub-indicator in the evaluation index set MI; k represents the weight ordinal number; Λ represents the evaluation index that must be met; α i are the evaluation indicators that must be met, except α i The sum of the weight coefficients of other evaluation indicators is 1.
6. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 5 is characterized in that: The step 43 is based on the evaluation index α that must be met i , remove those below α i The association rules are as follows: in, Represents association rules The relevance of.
7. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 6 is characterized in that: The step 44 is for a i The association rules are constructed, the corresponding extension association function is constructed, the multi-index extension association degree is obtained based on the extension association function, and the multi-index extension association degree is normalized. The specific process includes: Assume that the failure risk item set Z = {Z1, Z2…Z m } all meet the evaluation index α i , then the multi-index extension correlation degree is expressed as: K i =(K i (Z1),K i (Z2)…K i (Z m )),i=1,2,…n Where m represents the total number of failure risk items.
8. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 7 is characterized in that: The multi-index extensible correlation degree is standardized as follows: Among them, the standard correlation degree of each evaluation index of the fault risk item is: i =(k i1 ,k i2 …k im ),i=1,2…n.
9. The method for aircraft engine fault diagnosis based on extension association rule mining according to claim 8 is characterized in that: In step 4 and 5, the fault risk items are obtained based on the multi-index extensible correlation after normalization, and the fault risk items are sorted according to their merits. The item with the largest merit is the fault that needs to be processed first. The specific process includes: Among them, C1(Z j ) represents the weighted merit of each evaluation index, C2(Z j ) indicates that the goodness of each evaluation index takes the minimum value, C3(Z j ) indicates that the goodness of each evaluation index takes the maximum value; K(Z j ) represents the canonical correlation matrix; The failure risk item set Z={Z1,Z2…Z m } are sorted by their merits, and the item with the largest merit is the fault that needs to be handled first:
10. An aircraft engine fault diagnosis system based on extension association rule mining, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.
Citation Information
Patent Citations
Engine quality early warning service method and system, readable medium and electronic equipment
CN112528519A
Rapid early warning method for operation fault of vehicle-mounted printed circuit board assembly
CN113125937A
Electric vehicle charging equipment fault analysis and early warning method and system
CN113147476A
Anomaly event detection using frequent patterns
US11055405B1
