Method for diagnosing faults in an aeronautical equipment system based on rule-based reasoning and a knowledge base

Abstract

The application relates to a fault diagnosis method of an aviation equipment system based on rule reasoning and a knowledge base, which comprises the following steps: S1, constructing a fault diagnosis data model of the aviation equipment system, acquiring fault diagnosis information, and establishing a fault knowledge base; S2, determining associated fault diagnosis information of the aviation equipment system based on rule reasoning of a rule base, and acquiring a similarity result through the knowledge base; S3, analyzing the similarity result of the fault diagnosis information of the aviation equipment system, sorting, and outputting a fault diagnosis result. The application extracts and identifies fault features based on a keyword base; obtains matched associated LRUs based on rule reasoning, and records corresponding trace information; searches and matches fault diagnosis information based on the knowledge base, calculates information matching similarity, completes similarity sorting of a final fault result, and determines the fault diagnosis result of the aviation equipment system. The application improves the fault diagnosis efficiency of the aviation equipment system, reduces the fault diagnosis time, and prolongs the service cycle of the aviation equipment.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology for aviation equipment systems, and specifically relates to a fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base. Background Technology

[0002] As aviation equipment continues to upgrade and evolve, its performance and functions are constantly improving and expanding, leading to increasingly complex structures and more intricate fault phenomena and causes, posing a significant challenge to fault diagnosis and analysis.

[0003] Currently, fault diagnosis technologies for aviation equipment are mainly divided into two categories: traditional fault diagnosis technologies and intelligent fault diagnosis technologies. Traditional fault diagnosis technologies mainly rely on human experience and advanced instruments, which are time-consuming and labor-intensive, and have low diagnostic efficiency and accuracy. In particular, large and complex aviation equipment often exhibits complexity, uncertainty, and even multiple concurrent and random faults, making it very difficult to troubleshoot using traditional fault diagnosis methods. At the same time, with the continuous development of information technology, especially the continuous popularization and maturity of artificial intelligence technology, research on intelligent fault diagnosis technologies has emerged, making it possible to quickly and accurately diagnose faults in large and complex aviation equipment. However, these technologies generally have the following problems: (1) It is difficult to collect fault data for large aviation equipment systems, while the training of intelligent models requires massive amounts of fault data. (2) As the amount of data continues to increase, the computational efficiency of the model algorithm is not high. (3) It is difficult to evaluate the accuracy of the algorithm's diagnostic effect.

[0004] To address the shortcomings of existing technologies, this paper proposes a semi-intelligent fault diagnosis method based on rule bases and knowledge bases. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a fault diagnosis method for aviation equipment systems based on rule-based reasoning and a knowledge base. This invention identifies fault features through feature extraction from a keyword database; obtains matching associated LRUs based on rule-based reasoning and records their corresponding traceability information; performs fault diagnosis information search and matching based on the knowledge base, calculates information matching similarity, and finally ranks the fault results based on similarity to determine the fault diagnosis result of the aviation equipment system. This invention improves fault diagnosis efficiency, reduces fault diagnosis time, and extends the service life of aviation equipment.

[0006] To achieve the above objectives, the present invention discloses the following technical solution:

[0007] A fault diagnosis method for aviation equipment systems based on rule-based reasoning and knowledge base, comprising:

[0008] S1: Construct a fault diagnosis data model for aviation equipment systems, obtain fault diagnosis information, and build a fault knowledge base;

[0009] S11: Based on the aviation equipment product tree structure, create a four-level aviation equipment product tree, including model-level product tree, system-level product tree, LRU-level product tree, and LRM-level product tree.

[0010] S12: Based on the key fault terminology database structure of aviation equipment, create a key fault terminology database for aviation equipment systems, specifically including a fault feature terminology database, a phrase terminology database, an equivalent terminology database, and a synonym terminology database; the fault feature terminology database includes fault codes, functions, and fault phenomena constituted by LRU and LRM; the phrase terminology database includes fault feature values; the equivalent terminology database consists of words with the same meaning as phrase values ​​but with different representation methods; the synonym terminology database consists of synonyms of phrases.

[0011] S13: Based on the rule base data structure of aviation equipment, create a rule reasoning library for the aviation equipment system;

[0012] S14: Based on the knowledge base data structure of aviation equipment, create a knowledge base for the aviation equipment system, perform knowledge base search based on the rule reasoning results of step S13, obtain the final fault diagnosis information, calculate similarity values, and sort them;

[0013] S2: Based on the rule base, rule reasoning is used to determine the fault diagnosis information of the associated aviation equipment system, and similarity results are obtained through the knowledge base;

[0014] S21: Receive the input error text, use a word segmentation library to perform intelligent semantic word segmentation, and convert the input error text into error word groups. The specific sub-steps are as follows:

[0015] S211: Using the fault diagnosis information in step S1, construct a fault statistics dictionary and determine the prefix dictionary of fault words in the fault statistics dictionary;

[0016] S212: Using regular expressions, the input faulty text is segmented into clauses, using punctuation marks or non-Chinese characters as delimiters, to obtain the faulty input sentence; a directed acyclic graph is constructed for the faulty input sentence based on the prefix dictionary of faulty words;

[0017] S213: Use dynamic programming to find a path with the highest failure probability on the directed acyclic graph, and perform word segmentation based on this path; the method for obtaining the path with the highest failure probability is as follows:

[0018] ;

[0019] in, This is the path with the highest probability of failure in the aviation equipment system. Enter a sentence to describe a malfunction in an aviation equipment system. It is a function that takes the minimum value of the independent variable; A set of fault paths for fault words in an aviation equipment system; For the first The faulty word Fault path; For the first The faulty word Word frequency in the prefix dictionary; This is the sum of the frequencies of all faulty words in the prefix dictionary; Number the faulty words; This represents the total number of faulty words.

[0020] S214: Use dynamic programming algorithm to calculate the path with the highest failure probability and obtain the key failure phrases;

[0021] S22: Based on the key fault phrases output in step S21, search the synonym database obtained in step S12. Use the Robin-Karp algorithm to compare the key fault phrases with the information in the synonym database. Extract fault features from the successfully matched synonym database and construct a fault feature grouping dictionary fault map. Information from the synonym database that fails to match is directly discarded. The Robin-Karp algorithm calculates and compares the hash values ​​of strings. The specific calculation process is as follows:

[0022] ;

[0023] in, For strings The hash value; The parameter for comparing the first string; To identify the string to be matched; The ASCII code value corresponding to the character; This is the parameter for comparing the second string; Number the position of the string; For string distance numbering;

[0024] S23: Based on the fault feature grouping dictionary fault map in step S22, perform rule reasoning based on the rule base created in step S13, obtain the LRU correlation dictionary and record the corresponding tracing information, and construct the LRU correlation dictionary and fault tracing dictionary.

[0025] S24: Based on the LRU association dictionary in step S23, search the fault diagnosis knowledge base created in step S14 to obtain the similarity of fault diagnosis information associated with LRU names;

[0026] S25: Traverse the LRU association dictionary in step S23, use the fault tracking dictionary in step S23 to obtain the original input fault keywords corresponding to the current LRU name, and use the cosine similarity algorithm to calculate the similarity between the fault keywords and the fault diagnosis information of all LRU names in step S24.

[0027] S3: Analyze the similarity results of fault diagnosis information of aviation equipment systems, sort them and output the fault diagnosis results;

[0028] Based on the similarity results of the fault diagnosis information in step S2, the minimum heap algorithm is used to filter the fault diagnosis information by setting a similarity threshold, and the top n most similar fault diagnosis information are obtained. The similarity results of the filtered fault diagnosis information are sorted to obtain the fault diagnosis results of the aviation equipment system and output them.

[0029] Preferably, step S13, which involves creating a rule-based reasoning library for the aviation equipment system, includes the following sub-steps:

[0030] S131: Construct a fault code-LRU mapping rule base. Based on the fault code-LRU mapping rule base data structure described in the requirements specification, create a fault code-LRU mapping rule base to support rule reasoning based on fault codes and obtain associated fault code-LRU information, including fault codes and LRU names.

[0031] S132: Construct a function-LRU mapping rule base. Based on the function-LRU mapping rule base data structure described in the requirements specification, create a function-LRU mapping rule base to support function-based rule reasoning and obtain associated function-LRU information, including function name and LRU name.

[0032] S133: Construct a comprehensive mapping rule base. Based on the comprehensive rule-LRU mapping rule base data structure described in the requirements specification, create a comprehensive rule-LRU mapping rule base to support comprehensive rule reasoning based on the permutation and combination of functions, fault scenes, and fault codes, and obtain associated comprehensive LRU information, including the functional phenomenon combination name and LRU name.

[0033] Preferably, step S14 involves creating a knowledge base for the aviation equipment system, with the following specific steps:

[0034] S141: Construct a fault mode knowledge base; Based on the fault mode knowledge base data structure described in the requirements specification, create a fault mode knowledge base for the aviation equipment system, including fault codes, fault modes, and corresponding LRU names;

[0035] S142: Construct a testability knowledge base; Based on the testability knowledge base data structure described in the requirements specification, create a testability knowledge base for the aviation equipment system, including fault phenomena, fault codes, fault causes, fault LRMs, whether CNIs are reported, and LRU names;

[0036] S143: Construct a field fault knowledge base; Based on the field fault knowledge base data structure described in the requirements specification, create a field fault knowledge base for the aviation equipment system, including information such as recording time, aircraft type, factory code, fault phenomenon, fault cause, fault LRU, fault LRM, fault component type, fault code, discovery time, fault nature, fault malfunction code, identification method, and troubleshooting method.

[0037] S144: Construct a fault analysis manual knowledge base; Based on the fault analysis manual data structure described in the requirements specification, create a fault analysis manual knowledge base for the aviation equipment system, including: model, fault code, fault phenomenon, fault cause, fault LRU, fault LRM, and troubleshooting methods.

[0038] Preferably, in step S22, the fault feature grouping dictionary fault map is constructed, and the specific steps are as follows:

[0039] S221: Fault feature identification of function. The current fault keyword is compared with the records in the key fault word library whose fault feature is function. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0040] S222: Fault feature identification of fault codes. The current fault keyword is compared with the records in the key fault word library whose fault feature is fault code. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0041] S223: Fault feature identification: The current fault keyword is compared with the records in the key fault word library that are composed of fault features. If any field information of the record’s phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0042] S224: Fault feature identification of fault phenomena. The current fault keyword is compared with the records in the key fault term library whose fault feature is the fault phenomenon. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0043] Preferably, step S23 involves performing rule reasoning based on the rule base to obtain associated LRUs and record the corresponding tracing information. The specific steps are as follows:

[0044] S231: Create an LRU association dictionary as LruMap;

[0045] S232: Based on the aviation equipment system architecture, obtain the corresponding LRU information. The steps are as follows: Obtain the information of the fault feature grouping dictionary fault map that constitutes the LRM, and iterate through it to obtain the LRU information corresponding to the LRM from the aviation equipment system architecture. If it exists, first construct a first trace record composed of the LRU name, and then add the LRU name and the first trace record to the dictionary LruMap. The first trace record includes the LRU name and the word segmentation information corresponding to the LRM.

[0046] S233: Based on the fault code-LRU rule base in step S131, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the fault code information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed fault code with the fault codes in the rule base, and record the result as r; if r is successful, first construct the second trace record, and then add the LRU name and the second trace record to the dictionary LruMap; the second trace record includes the LRU name, fault code, and the word segmentation information corresponding to the fault code;

[0047] S234: Based on the function-LRU rule base in step S132, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the function information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed function with the function in the rule base, and record the result as r; if r is successful, first construct the third trace record, and then add the LRU name and the third trace record to the dictionary LruMap; the third trace record includes the LRU name, function, and the word segmentation information corresponding to the function.

[0048] S235: Based on the comprehensive rule mapping library in step S133, perform rule reasoning to obtain the corresponding LRU information and obtain the fault code FC, function FU and fault phenomenon PH in the fault feature grouping dictionary fault diagram.

[0049] Preferably, the similarity of fault diagnosis information in step S24 is as follows:

[0050] S241: Calculate the list length of the trace records for each LRU information in LruMap, and sort them in descending order according to the list length. The sorting result is: OrderLruMap.

[0051] S242: Retrieve the LRU names from OrderLruMap and construct a list of LRU names called LruNames;

[0052] S243: Based on the LRU name list LruNames, search the test test knowledge base created in step S142 to obtain the matching fault diagnosis information in the test test as: TestFaiLRUes; based on LruNames, search the field fault knowledge base created in step S143 to obtain the matching fault diagnosis information in the field fault knowledge base as: ForstageFaiLRUes; based on LruNames, search the fault analysis manual knowledge base created in step S144 to obtain the matching fault diagnosis information in the fault analysis manual knowledge base as: AnalysisManualFaiLRUes;

[0053] S244: Loop through OrderLruMap, calculate the similarity of the fault diagnosis information in step S243, and construct the final fault diagnosis result list as result; obtain the traceability information list in the current record of OrderLruMap as Traces;

[0054] S245: Using the cosine similarity algorithm in step S25, calculate the cosine similarity of fault diagnosis information in TestFaiLRUes, ForstageFaiLRUes, and AnlysisManualFaiLRUes;

[0055] S246: Based on the min-heap algorithm in step S3, the fault diagnosis result list result is filtered for similarity and sorted in descending order to form the final result.

[0056] Preferably, the cosine similarity algorithm in step S25 is as follows:

[0057] S251: Input fault keywords and LRU information corresponding to fault diagnosis information in the current knowledge base are set as the first fault text c1 and the second fault text c2, respectively; perform word segmentation operation and obtain word segmentation results; perform word frequency calculation on the word segmentation results and calculate the frequency of each word.

[0058] S252: Based on the result of step S251, the strings corresponding to the first fault text c1 and the second fault text c2 are respectively converted into word vectors composed of word frequencies as follows:

[0059] ;

[0060] in, The first faulty text is represented by its word frequency vector. The second fault text word frequency vector; , and These are the 1st, 2nd, and tth elements of the first fault text's word frequency vector, respectively. These are the 1st, 2nd, and tth elements of the word frequency vector of the second fault text, respectively.

[0061] S253: Calculate the similarity between the first faulty text c1 and the second faulty text c2:

[0062] ;

[0063] in, Similarity between texts; This is the k-th element of the word frequency vector of the first fault text; This is the k-th element of the second fault text word frequency vector; k is the element number of the fault text word frequency vector.

[0064] According to the cosine function, the more similar two texts are, the smaller the angle between their vectors.

[0065] S256: Normalize the similarity values ​​from step S255 so that the similarity values ​​are within the range of [0, 1], as shown in the following formula:

[0066] ;

[0067] in, The result is the normalized cosine similarity value.

[0068] Preferably, in step S3, the fault diagnosis information is filtered using a min-heap algorithm with a similarity threshold of 0.3, specifically as follows:

[0069] S31: Iterate through the fault diagnosis information with similarity scores until the min-heap contains 10 elements; set the similarity score of the currently iterated element to r1; when If the value is less than or equal to the minimum value, add it to the initial min-heap; otherwise, discard it. Adjust the min-heap by making the top element of the heap the minimum value, denoted as top. The min-heap is a binary tree structure, and the value of the parent node is smaller than the left and right child nodes, with the top of the heap being the minimum value.

[0070] S32: Iterate through the elements in the fault diagnosis information in turn. If the similarity of the elements is greater than 0.3 and greater than the top of the heap, remove the smallest element in the heap, add the current element, and perform heapification adjustment again so that the top element of the heap is the minimum value.

[0071] S33: Sort the elements in the min-heap in descending order and return the final fault diagnosis result.

[0072] Preferably, the fault code FC, function FU, and fault phenomenon PH in step S235 are as follows:

[0073] The information in the fault code FC is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a fourth trace record is constructed, and the LRU name and the fourth trace record are added to the dictionary LruMap. The fourth trace record consists of the LRU name, the fault code combination, and the word segmentation information corresponding to the fault code combination.

[0074] The information in the function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a fifth trace record is constructed, and the LRU name and the fifth trace record are added to the dictionary LruMap. The fifth trace record includes the LRU name, function combination, and word segmentation information corresponding to the function combination.

[0075] The information in the fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a sixth trace record is constructed, and the LRU name and the sixth trace record are added to the dictionary LruMap. The sixth trace record includes the LRU name, the fault phenomenon combination, and the word segmentation information corresponding to the fault phenomenon combination.

[0076] The information in the fault code FC and function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a seventh trace record is constructed, and the LRU name and the seventh trace record are added to the dictionary LruMap. The seventh trace record includes the LRU name, fault code, function combination, and word segmentation information corresponding to the fault code and function combination.

[0077] The information in the function FU and fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the eighth trace record is constructed, and the LRU name and the eighth trace record are added to the dictionary LruMap. The eighth trace record includes the LRU name, function, fault scene combination, fault code, and word segmentation information corresponding to the fault scene combination.

[0078] Preferably, step S245 calculates the similarity of fault diagnosis information in TestFaiLRUes, and the specific steps are as follows:

[0079] S2451: Based on the LRU name in OrderLruMap, retrieve all fault diagnosis information corresponding to the LRU association dictionary from TestFaiLRUes, denoted as CurrentFaiLRUes;

[0080] S2452: Loop through all trace records and record the information currently being traced as Trace;

[0081] S2453: If Trace is a fault code, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and fault codes in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0082] S2454: If Trace is a function, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and function information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0083] S2455: If the Trace is a fault phenomenon, the cosine similarity algorithm is used to calculate the similarity between the Trace segmentation information and the fault phenomenon information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0084] S2456: If Trace is a component, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and component information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0085] S2457: If the Trace is a comprehensive rule, then the comprehensive rule is split, and the similarity is calculated and accumulated according to the rules in steps S2451 to S2456. Finally, the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0086] Compared with the prior art, the present invention has the following beneficial effects:

[0087] (1) This invention solves the problem of difficulty in fault diagnosis caused by inconsistent feature terms by modeling the faults of aviation equipment systems into feature vectors composed of fault codes, functions, LRUs and LRMs, and fault phenomena, and constructing key fault terms. This invention achieves multi-dimensional fault diagnosis and greatly reduces the complexity of fault diagnosis.

[0088] (2) Compared with traditional fault diagnosis methods, this invention improves the accuracy of fault diagnosis and reduces the time of fault diagnosis by performing intelligent semantic recognition on fault input, extracting key feature information, and completing feature unification and weight calculation based on rule base.

[0089] (3) Compared with artificial intelligence-based fault diagnosis methods, this invention solves the problems of difficulty in collecting fault data of large and complex equipment, difficulty in supporting model training of artificial intelligence diagnostic algorithms, and large model generalization error by using rule-based reasoning and knowledge base search methods. This enhances the intelligence level of fault diagnosis and improves the efficiency of fault diagnosis.

[0090] (4) In the process of fault search in the knowledge base, the present invention introduces a fault feature text similarity algorithm to calculate the similarity of fault feature texts and sorts the similarity scores before presenting the final fault diagnosis results, thereby improving the accuracy and rationality of fault diagnosis. Attached Figure Description

[0091] Figure 1 This is a flowchart of the fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base of the present invention;

[0092] Figure 2 This is a schematic diagram of the system architecture of the fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base of the present invention.

[0093] Figure 3 This is an example diagram of the system configuration of the ICNC system in an embodiment of the present invention;

[0094] Figure 4 This is an example diagram of the key fault terminology library for a model ICNC system in an embodiment of the present invention;

[0095] Figure 5 This is an example diagram of the fault code-LRU mapping rule base of the ICNC system in an embodiment of the present invention;

[0096] Figure 6 This is an example diagram of the function-LRU mapping rule base of the model ICNC system in an embodiment of the present invention;

[0097] Figure 7 This is an example diagram of the integrated mapping rule base of the model ICNC system in an embodiment of the present invention;

[0098] Figure 8 This is an example diagram of the fault mode knowledge base of the ICNC system in an embodiment of the present invention.

[0099] Figure 9 This is an example diagram of the fault analysis manual knowledge base for the ICNC system in an embodiment of the present invention;

[0100] Figure 10 This is an example diagram of the test knowledge base for the model ICNC system in an embodiment of the present invention;

[0101] Figure 11 This is an example diagram of the field fault knowledge base of the model ICNC system in an embodiment of the present invention;

[0102] Figure 12 This is an example diagram showing the fault diagnosis results of the ICNC system in an embodiment of the present invention. Detailed Implementation

[0103] Exemplary embodiments, features, and aspects of the present invention will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0104] This invention provides a fault diagnosis method for aviation equipment systems based on rule-based reasoning and a knowledge base, such as... Figure 1 As shown, a fault diagnosis data model for aviation equipment systems is constructed, fault diagnosis information is obtained, a fault knowledge base is built, and a rule-based reasoning method is used to determine the similarity results of fault diagnosis information based on the rule base. The similarity results of fault diagnosis information of aviation equipment systems are analyzed, sorted, and the fault diagnosis results are output.

[0105] like Figure 2 This is a system architecture diagram of the fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base of the present invention. It consists of a client, a service layer, and a data layer. The client is mainly used to input information and display the output diagnostic results. The service layer is used to implement the core functions of fault feature extraction, rule reasoning, and knowledge base search. The data layer is mainly used to provide various rule bases and knowledge bases used by the present invention.

[0106] This invention takes the ICNC system of a certain aviation equipment system as a specific research object. Figure 3 This is an example diagram of the system configuration of the ICNC system in an embodiment of the present invention, which consists of 20 subsystems; its steps include:

[0107] Step S1: Construct a fault diagnosis data model for the aviation equipment system, obtain fault diagnosis information, and build a fault knowledge base;

[0108] Step S11: Based on the aviation equipment product tree structure in the aviation equipment specification, create a four-level aviation equipment product tree, specifically: model level, system level, LRU level, and LRM level.

[0109] Step S12: Based on the key fault terminology data structure in the specifications of the aviation equipment, create a key fault terminology database for the aviation equipment system, specifically including: a fault feature terminology database, a phrase terminology database, an equivalent terminology database, and a synonym terminology database; such as Figure 4 The diagram shows an example of a key fault terminology library for a model ICNC system in an embodiment of the present invention. The fault feature terminology library types include: fault codes, functions, LRU and LRM configurations, and fault phenomena. The phrase terminology library describes fault feature values. The synonym terminology library contains words with the same meaning as the phrase values ​​but representing different terms. The near-synonym terminology library contains near-synonyms of the phrases.

[0110] Step S13: Based on the rule base data structure of the aviation equipment specification, create a rule inference base for the aviation equipment system. The specific steps are as follows:

[0111] Step S131: Construct a fault code-LRU mapping rule base, such as Figure 5 The diagram shows an example of the fault code-LRU mapping rule base for the ICNC system in the embodiment of the present invention. Based on the fault code-LRU mapping rule base data structure in the requirements specification, a fault code-LRU mapping rule base is created to support rule reasoning based on fault codes and to obtain associated fault code-LRU information, including: fault code and LRU name.

[0112] Step S132: Construct a feature-LRU mapping rule base, such as Figure 6 The diagram shows an example of the function-LRU mapping rule base of the ICNC system in the embodiment of the present invention. According to the function-LRU mapping rule base data structure in the requirements specification, a function-LRU mapping rule base is created to support function-based rule reasoning and obtain associated function-LRU information, including: function name and LRU name.

[0113] Step S133: Construct a comprehensive mapping rule base, such as Figure 7 The diagram shows an example of the integrated mapping rule base of the ICNC system in the embodiment of the present invention. According to the integrated rule-LRU mapping rule base data structure in the requirements specification, an integrated rule-LRU mapping rule base is created to support integrated rule reasoning based on the permutation and combination of functions, fault scenes, and fault codes, and to obtain the associated integrated LRU information, including: the name of the functional phenomenon combination and the LRU name.

[0114] Step S14: Based on the knowledge base data structure of the aviation equipment specifications, create a knowledge base for the aviation equipment system. This involves searching the knowledge base based on the rule-based reasoning results from step S13, obtaining the final fault diagnosis information, calculating similarity values, and sorting them to build the fault knowledge base. The specific steps for creating the knowledge base for the aviation equipment system are as follows:

[0115] Step S141: Construct a fault mode knowledge base, such as Figure 8 The diagram illustrates an example of the fault mode knowledge base of the ICNC system in an embodiment of the present invention. Based on the fault mode knowledge base data structure described in the requirements specification, a fault mode knowledge base for an aviation equipment system is created, including: fault codes, fault modes, and corresponding LRU names.

[0116] Step S142: Construct a knowledge base for test experiments, such as Figure 10 The diagram illustrates a test knowledge base for the ICNC system of the present invention. Based on the test knowledge base data structure described in the requirements specification, a test knowledge base for the aviation equipment system is created, including: fault phenomena, fault codes, fault causes, fault LRMs, whether CNIs are reported, and LRU names.

[0117] Step S143: Construct an outdoor fault knowledge base, such as... Figure 11 The diagram illustrates an example of the field fault knowledge base for the ICNC system in an embodiment of the present invention. Based on the field fault knowledge base data structure described in the requirements specification, a field fault knowledge base for an aviation equipment system is created, including: recording time, specialty, aircraft type, factory code, fault phenomenon, fault cause, fault LRU, fault LRM, fault component type, fault code, discovery time, fault nature, fault malfunction code, identification method, and troubleshooting method.

[0118] Step S144: Construct a fault analysis manual knowledge base, such as Figure 9 The diagram illustrates an example of a fault analysis manual knowledge base for an ICNC system in an embodiment of the present invention. Based on the fault analysis manual data structure described in the requirements specification, a fault analysis manual knowledge base for an aviation equipment system is created, including: model, fault code, fault phenomenon, fault cause, fault LRU, fault LRM, and troubleshooting method.

[0119] Step S2: Determine the similarity results of fault diagnosis information of the aviation equipment system based on rule reasoning using the rule base;

[0120] Step S21: Receive the input fault text "0x48121001 Data link attitude information disappeared". Use a word segmentation library to complete intelligent semantic word segmentation of the text, converting the input fault text "E0001 Pilot's voice" into the key fault phrases [E0001, Pilot's voice]. The specific steps are as follows:

[0121] Step S211: Construct a fault statistics dictionary using the fault diagnosis information from step S1, and determine the prefix dictionary of fault words in the fault statistics dictionary.

[0122] Step S212: Using regular expressions, the input fault text is segmented into clauses, with punctuation marks or non-Chinese characters as delimiters, to obtain the fault input sentence; a directed acyclic graph is constructed for the fault input sentence based on the prefix dictionary of fault words.

[0123] Step S213: Use dynamic programming to find the path with the highest failure probability on the directed acyclic graph, and perform word segmentation based on this path.

[0124] First, the model for constructing the path with the highest failure probability is as follows:

[0125] ;

[0126] in, This is the path with the highest probability of failure. Enter a sentence for the fault; It is a function that takes the minimum value of the independent variable; This is the set of fault paths for faulty words.

[0127] Then, because the faulty input sentence is independent, we get

[0128] ;

[0129] in, For the first The faulty word Fault path; This is the first faulty word; This is the second faulty word; This is the nth faulty word; Number the faulty words; This represents the total number of faulty words.

[0130] Secondly The faulty word The fault path is obtained according to the following formula:

[0131] ;

[0132] in, For the first The faulty word Word frequency in the prefix dictionary; This is the sum of the frequencies of all faulty words in the prefix dictionary;

[0133] The final method for obtaining the path with the highest failure probability is as follows:

[0134] ;

[0135] Step S214: Use dynamic programming algorithm to calculate the path with the highest failure probability and obtain the key failure phrases;

[0136] Step S22: Based on the key fault phrases output in Step S21, search the synonym database obtained in Step S12. Use the Robin-Karp algorithm to compare the key fault phrases with the information in the synonym database. Extract fault features from the successfully matched synonym database. The matching result for "E0001" is: associated fault codes 0x4812E0001 and 0x4822E0001. Match the fault features of the "pilot's voice" function. Construct a fault feature grouping dictionary fault map. Discard information from the synonym database that failed to match.

[0137] The Robin-Karp algorithm calculates and compares the hash values ​​of strings. The specific calculation process is as follows:

[0138] ;

[0139] in, For strings The hash value; The parameter for comparing the first string; To identify the string to be matched; The ASCII code value corresponding to the character; This is the parameter for comparing the second string; Number the position of the string; This is the distance number for the string.

[0140] The specific steps for constructing a fault feature grouping dictionary and fault map are as follows:

[0141] Step S221: Functional fault feature identification. The current fault keyword is compared with records in the key fault terminology database whose fault feature is "function". If any field of information in the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram. Here, "E0001" and "pilot's voice" have no matching near-synonyms or equivalent words.

[0142] Step S222: Fault feature identification of fault codes. The current fault keyword is compared with the records in the key fault term library whose fault feature is fault code. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0143] Step S223: Fault feature identification. The current fault keyword is compared with the records in the key fault term library that are composed of fault features. If any field information of the record’s phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0144] Step S224: Fault feature identification of fault phenomenon. The current fault keyword is compared with the records in the key fault term library whose fault feature is fault phenomenon. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

[0145] After the above steps, the content of the fault feature grouping dictionary fault diagram is: [Fault Code: [0x4812E0001, 0x4822E0001], Function: [Pilot Voice]].

[0146] Step S23: Based on the fault feature grouping dictionary and fault map from step S22, perform rule reasoning based on the rule base created in step S13, obtain the LRU correlation dictionary and record the corresponding tracing information, and construct the LRU correlation dictionary and fault tracing dictionary. The specific steps for performing rule reasoning based on the rule base to obtain associated LRUs and record the corresponding tracing information are as follows:

[0147] Step S231: Create an LRU association dictionary called LruMap;

[0148] Step S232: Based on the aviation equipment system architecture, obtain the corresponding LRU information. The steps are as follows: Obtain the LRM information of the fault feature grouping dictionary fault map, and iterate through it to obtain the LRU information corresponding to the LRM from the aviation equipment system architecture. If it exists, first construct the first trace record, and then add the LRU name and the first trace record to the dictionary LruMap. The first trace record includes: LRU name and word segmentation information corresponding to LRM.

[0149] Step S233: Based on the fault code-LRU rule base in step S131, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the fault code information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed fault code with the fault codes in the rule base, and record the result as r; if r is successful, first construct the second trace record, and then add the LRU name and the second trace record to the dictionary LruMap; the second trace record includes: LRU name, fault code, and word segmentation information corresponding to the fault code;

[0150] Step S234: Based on the function-LRU rule base in step S132, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the function information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed function with the function in the rule base, and record the result as r; if r is successful, first construct the third trace record, and then add the LRU name and the third trace record to the dictionary LruMap; the third trace record includes: LRU name, function, and word segmentation information corresponding to the function;

[0151] Step S235: Perform rule reasoning based on the comprehensive rule mapping library in step S133, obtain the corresponding LRU information, and obtain the fault code FC, function FU, and fault phenomenon PH from the fault feature grouping dictionary fault diagram.

[0152] The information in the fault code FC is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a fourth trace record is constructed, and the LRU name and the fourth trace record are added to the dictionary LruMap. The fourth trace record consists of the LRU name, the fault code combination, and the word segmentation information corresponding to the fault code combination.

[0153] The information in the function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the fifth trace record is constructed, and the LRU name and the fifth trace record are added to the dictionary LruMap. The fifth trace record consists of the LRU name, function combination, and word segmentation information corresponding to the function combination.

[0154] The information in the fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the sixth trace record is constructed, and the LRU name and the sixth trace record are added to the dictionary LruMap. The sixth trace record consists of the LRU name, the fault phenomenon combination, and the word segmentation information corresponding to the fault phenomenon combination.

[0155] The information in the fault code FC and function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the seventh trace record is constructed, and the LRU name and the seventh trace record are added to the dictionary LruMap. The seventh trace record consists of the LRU name, fault code, function combination, fault code, and word segmentation information corresponding to the function combination.

[0156] The information in the function FU and fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the eighth trace record is constructed, and the LRU name and the eighth trace record are added to the dictionary LruMap. The eighth trace record consists of the LRU name, function, fault scene combination, fault code, and word segmentation information corresponding to the fault scene combination.

[0157] Step S24: Based on the LRU association dictionary in step S23, search the fault diagnosis knowledge base created in step S14 to obtain the similarity of fault diagnosis information associated with LRU names. The specific steps are as follows:

[0158] Step S241: Calculate the list length of the trace records for each LRU information in LruMap, and sort them in descending order according to the list length. The sorting result is: OrderLruMap.

[0159] Step S242: Obtain the LRU names in OrderLruMap and construct a list of LRU names called LruNames.

[0160] Step S243: Based on the LRU name list LruNames, search the test test knowledge base created in step S142 to obtain the matching fault diagnosis information in the test test as: TestFaiLRUes; based on LruNames, search the field fault knowledge base created in step S143 to obtain the matching fault diagnosis information in the field fault knowledge base as: ForstageFaiLRUes; based on LruNames, search the fault analysis manual knowledge base created in step S144 to obtain the matching fault diagnosis information in the fault analysis manual knowledge base as: AnalysisManualFaiLRUes.

[0161] Step S244: Iterate through OrderLruMap, calculate the similarity of the fault diagnosis information in step S243, and construct the final fault diagnosis result list as result; obtain the traceability information list in the current record of OrderLruMap as Traces.

[0162] Step S245: Using the cosine similarity algorithm from step S25, calculate the similarity of fault diagnosis information in TestFaiLRUes. The specific steps are as follows:

[0163] Step S2451: Based on the LRU name in OrderLruMap, retrieve all fault diagnosis information corresponding to the LRU association dictionary from TestFaiLRUes, denoted as CurrentFaiLRUes.

[0164] Step S2452: Loop through all trace records and record the information being traced as Trace.

[0165] Step S2453: If Trace is a fault code, the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and fault codes in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0166] Create variables key and value; using "0x4812E0001" and "0x4822E0001" as conditions, search the fault code rule base sequentially, and the matched LRUs are

Integrated Channel Extension 1

Integrated Channel Extension 2

[0167] Step S2454: If Trace is a function, then use the cosine similarity algorithm to calculate the similarity between Trace segmentation information and function information in CurrentFaiLRUes, and combine the maximum similarity with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0168] Create variables key and value; using "pilot voice" as the condition, search the function rule base sequentially, and the LRU that matches is "pilot voice"; construct key-value pairs of {pilot voice, [pilot voice, pilot voice, backup control box, function]} and add them to LruMap.

[0169] Step S2455: If the Trace is a fault phenomenon, then use the cosine similarity algorithm to calculate the similarity between the Trace segmentation information and the fault phenomenon information in CurrentFaiLRUes, and combine the maximum similarity with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. Since there is no fault phenomenon information here, this step is unnecessary.

[0170] Step S2456: If the Trace is a component, then use the cosine similarity algorithm to calculate the similarity between the Trace segmentation information and the component information in CurrentFaiLRUes, and combine the maximum similarity with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. Since there is no component information here, this step is unnecessary.

[0171] Step S2457: If the Trace is a comprehensive rule, then the comprehensive rule is split, and the similarity is calculated and accumulated according to the rules in steps S2451 to S2456. Finally, the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

[0172] Step S246: Based on the min-heap algorithm in step S3, the fault diagnosis result list result is filtered for similarity and sorted in descending order to form the final result.

[0173] The calculation method for the cosine similarity of fault diagnosis information in ForstageFaiLRUes and AnlysisManualFaiLRUes is similar.

[0174] The LruMap results are as follows:

{Integrated Channel Extension 1, [E001, 0x4812E0001, Integrated Channel Extension 1, Fault Code]}, {Integrated Channel Extension 2, [E001, 0x4822E0001, Integrated Channel Extension 2, Fault Code]}, {Pilot Voice, [Pilot Voice, Pilot Voice, Backup Control Box, Function]}

[0175] Step S25: Traverse the LRU association dictionary in Step S23, obtain the original input fault keywords corresponding to the current LRU name using the fault tracking dictionary in Step S23, and calculate the similarity between the fault keywords and the fault diagnosis information of all LRU names in Step S24 using the cosine similarity algorithm; the cosine similarity algorithm is as follows:

[0176] Step S251: Input fault keywords and LRU information corresponding to fault diagnosis information in the current knowledge base, set as first fault text c1 and second fault text c2 respectively; perform word segmentation operation and obtain word segmentation results; perform word frequency calculation on the word segmentation results and calculate the frequency of each word.

[0177] Step S252: Based on the result of step S251, the strings corresponding to the first fault text c1 and the second fault text c2 are converted into word vectors composed of word frequencies, respectively:

[0178] ;

[0179] in, The first faulty text is represented by its word frequency vector. The second fault text word frequency vector; , and These are the 1st, 2nd, and tth elements of the first fault text's word frequency vector, respectively. These are the 1st, 2nd, and tth elements of the word frequency vector of the second fault text, respectively.

[0180] Step S253: Calculate the similarity between the first faulty text c1 and the second faulty text c2:

[0181] ;

[0182] in, Similarity between texts; This is the k-th element of the word frequency vector of the first fault text; This is the k-th element of the second fault text word frequency vector; k is the element number of the fault text word frequency vector.

[0183] According to the cosine function, the more similar two texts are, the smaller the angle between their vectors.

[0184] Step S256: Normalize the similarity values ​​from step S255 so that the similarity values ​​are within the range of [0, 1], as shown in the following formula:

[0185] ;

[0186] in, The result is the normalized cosine similarity value.

[0187] S3: Analyze the similarity results of fault diagnosis information of the aviation equipment system, sort them, and output the fault diagnosis results; based on the similarity results of fault diagnosis information in step S2, use the min-heap algorithm to filter the fault diagnosis information by setting a similarity threshold, obtain the top n most similar fault diagnosis information, sort the similarity results of the filtered fault diagnosis information, obtain the fault diagnosis results of the aviation equipment system, and output them. The similarity threshold is set to 0.3 to filter the fault diagnosis information and obtain the top n most similar ones, where n is 10.

[0188] Specifically:

[0189] Step S31: Iterate through the fault diagnosis information with similarity scores until the min-heap contains 10 elements; set the similarity score of the currently iterated element to r1; when If the value is less than or equal to the minimum value, it is added to the initial min-heap; otherwise, it is discarded. The min-heap is then heapified so that the top element of the heap is the minimum value, denoted as top. Therefore, a min-heap is a binary tree structure in which the value of the parent node is smaller than the values ​​of its left and right child nodes, and the top of the heap is the minimum value.

[0190] Step S32: Iterate through the elements in the fault diagnosis information. If the similarity of an element is greater than 0.3 and greater than the top of the heap, remove the smallest element from the heap, add it to the current element, and perform heapification adjustment again so that the top element of the heap is the minimum value.

[0191] Step S33: Sort the elements in the min-heap in descending order and return the final fault diagnosis result.

[0192] Sort the LruMap in descending order based on the number of values, and denote the result as LruOrderMap. Search the knowledge base based on the key values ​​(i.e., Lru names) in LruOrderMap to obtain matching results. The specific steps are as follows: Obtain the key values ​​of LruOrderMap, and construct the following query conditions: [Integrated Channel Extension 1, Integrated Channel Extension 2, Pilot Voice]. Using the above list as the query conditions, search the Fault Analysis Manual knowledge base, and denote the result as result1. Using the above list as the query conditions, search the Test Experiment knowledge base, and denote the result as result2. Using the above list as the query conditions, search the Field Fault Data knowledge base, and denote the result as result3.

[0193] Traverse the LruOrderMap and perform similarity calculations. The specific steps are as follows: Construct a storage list for the final fault diagnosis results, denoted as result. If the currently traversed record is {Integrated Channel Extension 1, [E001, 0x4812E0001, Integrated Channel Extension 1, Fault Code]} or {Integrated Channel Extension 2, [E001, 0x4822E0001, Integrated Channel Extension 2, Fault Code]}, then perform the following steps: Obtain the fault data corresponding to "Integrated Channel Extension 1" from result1, denoted as result1_failures. Traverse the fault data in result1_failures, calculate the cosine similarity value between the fault code of each record and "E0001", and construct a new record composed of the current fault and the cosine similarity value, storing it in result. If the currently traversed record is {Pilot Voice, [Pilot Voice, Pilot Voice, Backup Control Box, Function]}, then perform the following steps: Obtain the fault data corresponding to "Pilot Voice" from result2, denoted as result2_failures. Traverse the fault data in result2_failures, calculate the cosine similarity value between the function and "Pilot Voice" of each record, and construct a new record composed of the current fault and the cosine similarity value, storing it in result. Since there are no fault phenomena, composition, or functional fault combination characteristics, there is no need to perform the corresponding similarity calculation.

[0194] Sort the results according to their similarity scores, such as... Figure 12The figure shown is an example diagram of the fault diagnosis result of the model ICNC system in an embodiment of the present invention; the most similar n with a similarity threshold of sv and above are obtained. In a preferred embodiment, the value of parameter sv in the fault diagnosis information of the aviation equipment system is 30% and the value of n is 10. The similarity values ​​are sorted according to their size to obtain the fault diagnosis result of the aviation equipment system and output it.

[0195] The beneficial effects of this invention are as follows: This invention provides a fault diagnosis method for aviation equipment systems based on rule-based reasoning and a knowledge base. Taking the ICNC system as a specific research object, this invention solves the problem of difficulty in fault diagnosis caused by inconsistent feature terminology by modeling the faults of the ICNC system, achieving multi-dimensional fault diagnosis and reducing the complexity of fault diagnosis. Compared with traditional fault diagnosis methods, this invention improves the accuracy of fault diagnosis and reduces the time required by intelligently recognizing the fault input, extracting key fault feature information, and completing feature recognition based on a rule base. Compared with artificial intelligence-based fault diagnosis methods, this invention enhances the intelligence level of fault diagnosis and improves the efficiency of fault diagnosis through rule-based reasoning and knowledge base search methods. During the fault search process in the knowledge base, this invention introduces a fault feature text similarity algorithm to calculate the similarity of fault feature texts and rank the similarity scores. Through the analysis and demonstration of the results of this invention in the embodiments, this invention's method can improve the accuracy and rationality of fault diagnosis.

[0196] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A fault diagnosis method for aviation equipment systems based on rule-based reasoning and knowledge base, characterized in that, It includes: S1: Construct a fault diagnosis data model for aviation equipment systems, obtain fault diagnosis information, and build a fault knowledge base; S11: Based on the aviation equipment product tree structure, create a four-level aviation equipment product tree, including model-level product tree, system-level product tree, LRU-level product tree, and LRM-level product tree. S12: Based on the key fault terminology database structure of aviation equipment, create a key fault terminology database for aviation equipment systems, specifically including a fault feature terminology database, a phrase terminology database, an equivalent terminology database, and a synonym terminology database; the fault feature terminology database includes fault codes, functions, and fault phenomena constituted by LRU and LRM; the phrase terminology database includes fault feature values; the equivalent terminology database consists of words with the same meaning as phrase values ​​but with different representation methods; the synonym terminology database consists of synonyms of phrases. S13: Based on the rule base data structure of aviation equipment, create a rule reasoning library for the aviation equipment system; S14: Based on the knowledge base data structure of aviation equipment, create a knowledge base for the aviation equipment system, perform knowledge base search based on the rule reasoning results of step S13, obtain the final fault diagnosis information, calculate similarity values, and sort them; S2: Based on the rule base, rule reasoning is used to determine the fault diagnosis information of the associated aviation equipment system, and similarity results are obtained through the knowledge base; S21: Receive the input fault text, use a word segmentation library to perform intelligent semantic word segmentation, and convert the input fault text into key fault phrases. The specific sub-steps are as follows: S211: Using the fault diagnosis information in step S1, construct a fault statistics dictionary and determine the prefix dictionary of fault words in the fault statistics dictionary; S212: Using regular expressions, the input faulty text is segmented into clauses, using punctuation marks or non-Chinese characters as delimiters, to obtain the faulty input sentence; a directed acyclic graph is constructed for the faulty input sentence based on the prefix dictionary of faulty words; S213: Use dynamic programming to find a path with the highest failure probability on the directed acyclic graph, and perform word segmentation based on this path; the method for obtaining the path with the highest failure probability is as follows: ； in, This is the path with the highest probability of failure in the aviation equipment system. Enter a sentence to describe a malfunction in an aviation equipment system. It is a function that takes the minimum value of the independent variable; A set of fault paths for fault words in an aviation equipment system; For the first One faulty word Fault path; For the first One faulty word Word frequency in the prefix dictionary; This is the sum of the frequencies of all faulty words in the prefix dictionary; Number the faulty words; This represents the total number of faulty words. S214: Use dynamic programming algorithm to calculate the path with the highest failure probability and obtain the key failure phrases; S22: Based on the key fault phrases output in step S21, search the synonym database obtained in step S12. Use the Robin-Karp algorithm to compare the key fault phrases with the information in the synonym database. Extract fault features from the successfully matched synonym database and construct a fault feature grouping dictionary fault map. Information from the synonym database that fails to match is directly discarded. The Robin-Karp algorithm calculates and compares the hash values ​​of strings. The specific calculation process is as follows: ； in, For string The hash value; The parameter for comparing the first string; To identify the string to be matched; The ASCII code value corresponding to the character; This is the parameter for comparing the second string; Number the position of the string; For string distance numbering; S23: Based on the fault feature grouping dictionary fault map in step S22, perform rule reasoning based on the rule base created in step S13, obtain the LRU correlation dictionary and record the corresponding tracing information, and construct the LRU correlation dictionary and fault tracing dictionary. S24: Based on the LRU association dictionary in step S23, search the fault diagnosis knowledge base created in step S14 to obtain the similarity of fault diagnosis information associated with LRU names; S25: Traverse the LRU association dictionary in step S23, use the fault tracking dictionary in step S23 to obtain the original input fault keywords corresponding to the current LRU name, and use the cosine similarity algorithm to calculate the similarity between the fault keywords and the fault diagnosis information of all LRU names in step S24. S3: Analyze the similarity results of fault diagnosis information of aviation equipment systems, sort them and output the fault diagnosis results; Based on the similarity results of the fault diagnosis information in step S2, the minimum heap algorithm is used to filter the fault diagnosis information by setting a similarity threshold, and the top n most similar fault diagnosis information are obtained. The similarity results of the filtered fault diagnosis information are sorted to obtain the fault diagnosis results of the aviation equipment system and output them.

2. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: Step S13, creating the rule reasoning library for the aviation equipment system, specifically includes the following sub-steps: S131: Construct a fault code-LRU mapping rule base. Based on the fault code-LRU mapping rule base data structure described in the requirements specification, create a fault code-LRU mapping rule base to support rule reasoning based on fault codes and obtain associated fault code-LRU information, including fault codes and LRU names. S132: Construct a function-LRU mapping rule base. Based on the function-LRU mapping rule base data structure described in the requirements specification, create a function-LRU mapping rule base to support function-based rule reasoning and obtain associated function-LRU information, including function name and LRU name. S133: Construct a comprehensive mapping rule base. Based on the comprehensive rule-LRU mapping rule base data structure described in the requirements specification, create a comprehensive rule-LRU mapping rule base to support comprehensive rule reasoning based on the permutation and combination of functions, fault scenes, and fault codes, and obtain associated comprehensive LRU information, including the functional phenomenon combination name and LRU name.

3. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: Step S14 involves creating a knowledge base for the aviation equipment system. The specific steps are as follows: S141: Construct a fault mode knowledge base; Based on the fault mode knowledge base data structure described in the requirements specification, create a fault mode knowledge base for the aviation equipment system, including fault codes, fault modes, and corresponding LRU names; S142: Construct a knowledge base for test experiments; Based on the testability test knowledge base data structure described in the requirements specification, create a testability test knowledge base for the aviation equipment system, including fault phenomena, fault codes, fault causes, fault LRMs, whether CNIs are reported, and LRU names. S143: Construct an outdoor fault knowledge base; Based on the field fault knowledge base data structure described in the requirements specification, create a field fault knowledge base for the aviation equipment system, including information such as recording time, aircraft type, factory code, fault phenomenon, fault cause, fault LRU, fault LRM, fault component type, fault code, discovery time, fault nature, fault malfunction code, identification method, and troubleshooting method. S144: Construct a knowledge base for fault analysis manuals; Based on the fault analysis manual data structure described in the requirements specification, create a fault analysis manual knowledge base for aviation equipment systems, including: model, fault code, fault phenomenon, fault cause, fault LRU, fault LRM, and troubleshooting methods.

4. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: Step S22 involves constructing a fault feature grouping dictionary fault map. The specific steps are as follows: S221: Fault feature identification of function. The current fault keyword is compared with the records in the key fault word library whose fault feature is function. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram. S222: Fault feature identification of fault codes. The current fault keyword is compared with the records in the key fault word library whose fault feature is fault code. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram. S223: Fault feature identification: The current fault keyword is compared with the records in the key fault word library that are composed of fault features. If any field information of the record’s phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram. S224: Fault feature identification of fault phenomena. The current fault keyword is compared with the records in the key fault term library whose fault feature is the fault phenomenon. If any field information of the record's phrase, equivalent word, or near-synonym matches the fault keyword, the comparison is successful, and the corresponding phrase information is added to the fault feature grouping dictionary fault diagram.

5. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: Step S23, which involves performing rule reasoning based on the rule base, obtaining associated LRUs, and recording the corresponding tracing information, includes the following steps: S231: Create an LRU association dictionary as LruMap; S232: Based on the aviation equipment system architecture, obtain the corresponding LRU information. The steps are as follows: Obtain the information of the fault feature grouping dictionary fault map that constitutes the LRM, and iterate through it to obtain the LRU information corresponding to the LRM from the aviation equipment system architecture. If it exists, first construct a first trace record composed of the LRU name, and then add the LRU name and the first trace record to the dictionary LruMap. The first trace record includes the LRU name and the word segmentation information corresponding to the LRM. S233: Based on the fault code-LRU rule base in step S131, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the fault code information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed fault code with the fault codes in the rule base, and record the result as r; if r is successful, first construct the second trace record, and then add the LRU name and the second trace record to the dictionary LruMap; the second trace record includes the LRU name, fault code, and the word segmentation information corresponding to the fault code; S234: Based on the function-LRU rule base in step S132, perform rule reasoning to obtain the corresponding LRU information. The steps are as follows: Obtain the function information in the fault feature grouping dictionary fault map and iterate through it; compare the content of the currently traversed function with the function in the rule base, and record the result as r; if r is successful, first construct the third trace record, and then add the LRU name and the third trace record to the dictionary LruMap; the third trace record includes the LRU name, function, and the word segmentation information corresponding to the function. S235: Based on the comprehensive rule mapping library in step S133, perform rule reasoning to obtain the corresponding LRU information and obtain the fault code FC, function FU and fault phenomenon PH in the fault feature grouping dictionary fault diagram.

6. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: The similarity of fault diagnosis information in step S24 is as follows: S241: Calculate the list length of the trace records for each LRU information in LruMap, and sort them in descending order according to the list length. The sorting result is: OrderLruMap. S242: Retrieve the LRU names from OrderLruMap and construct a list of LRU names called LruNames; S243: Based on the LRU name list LruNames, search the test test knowledge base created in step S142 to obtain the matching fault diagnosis information in the test test as: TestFaiLRUes; based on LruNames, search the field fault knowledge base created in step S143 to obtain the matching fault diagnosis information in the field fault knowledge base as: ForstageFaiLRUes; based on LruNames, search the fault analysis manual knowledge base created in step S144 to obtain the matching fault diagnosis information in the fault analysis manual knowledge base as: AnalysisManualFaiLRUes; S244: Loop through OrderLruMap, calculate the similarity of the fault diagnosis information in step S243, and construct the final fault diagnosis result list as result; obtain the traceability information list in the current record of OrderLruMap as Traces; S245: Using the cosine similarity algorithm in step S25, calculate the cosine similarity of fault diagnosis information in TestFaiLRUes, ForstageFaiLRUes, and AnlysisManualFaiLRUes; S246: Based on the min-heap algorithm in step S3, the fault diagnosis result list result is filtered for similarity and sorted in descending order to form the final result.

7. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: The cosine similarity algorithm in step S25 is as follows: S251: Input fault keywords and LRU information corresponding to fault diagnosis information in the current knowledge base are set as the first fault text c1 and the second fault text c2, respectively; perform word segmentation operation and obtain word segmentation results; perform word frequency calculation on the word segmentation results and calculate the frequency of each word. S252: Based on the result of step S251, the strings corresponding to the first fault text c1 and the second fault text c2 are respectively converted into word vectors composed of word frequencies as follows: ； in, The first faulty text is represented by its word frequency vector. The second fault text word frequency vector; , and These are the 1st, 2nd, and tth elements of the first fault text's word frequency vector, respectively. These are the 1st, 2nd, and tth elements of the word frequency vector of the second fault text, respectively. S253: Calculate the similarity between the first faulty text c1 and the second faulty text c2: ； in, Similarity between texts; This is the k-th element of the word frequency vector of the first fault text; This is the k-th element of the second fault text word frequency vector; k is the element number of the fault text word frequency vector. According to the cosine function, the more similar two texts are, the smaller the angle between their vectors. S256: Normalize the similarity values ​​from step S255 so that the similarity values ​​are within the range of [0, 1], as shown in the following formula: ； in, The result is the normalized cosine similarity value.

8. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 1, characterized in that: In step S3, the minimum heap algorithm is used to filter the fault diagnosis information, with a similarity threshold of 0.

3. Specifically: S31: Iterate through the fault diagnosis information with similarity scores until the min-heap contains 10 elements; set the similarity score of the currently iterated element to r1; when If the value is less than or equal to the minimum value, add it to the initial min-heap; otherwise, discard it. Adjust the min-heap by making the top element of the heap the minimum value, denoted as top. The min-heap is a binary tree structure, and the value of the parent node is smaller than the left and right child nodes, with the top of the heap being the minimum value. S32: Iterate through the elements in the fault diagnosis information in turn. If the similarity of the elements is greater than 0.3 and greater than the top of the heap, remove the smallest element in the heap, add it to the current element, and perform heapification adjustment again so that the top element of the heap is the minimum value. S33: Sort the elements in the min-heap in descending order and return the final fault diagnosis result.

9. The fault diagnosis method for aviation equipment systems based on rule-based reasoning and knowledge base according to claim 5, characterized in that: The fault codes FC, FU, and PH in step S235 are as follows: The information in the fault code FC is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a fourth trace record is constructed, and the LRU name and the fourth trace record are added to the dictionary LruMap. The fourth trace record consists of the LRU name, the fault code combination, and the word segmentation information corresponding to the fault code combination. The information in the function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a fifth trace record is constructed, and the LRU name and the fifth trace record are added to the dictionary LruMap. The fifth trace record includes the LRU name, function combination, and word segmentation information corresponding to the function combination. The information in the fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a sixth trace record is constructed, and the LRU name and the sixth trace record are added to the dictionary LruMap. The sixth trace record includes the LRU name, the fault phenomenon combination, and the word segmentation information corresponding to the fault phenomenon combination. The information in the fault code FC and function FU is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, a seventh trace record is constructed, and the LRU name and the seventh trace record are added to the dictionary LruMap. The seventh trace record includes the LRU name, fault code, function combination, and word segmentation information corresponding to the fault code and function combination. The information in the function FU and fault phenomenon PH is arranged and combined, and the combined information is compared with the information in the comprehensive rule mapping library. If the comparison is successful, the eighth trace record is constructed, and the LRU name and the eighth trace record are added to the dictionary LruMap. The eighth trace record includes the LRU name, function, fault scene combination, fault code, and word segmentation information corresponding to the fault scene combination.

10. The fault diagnosis method for aviation equipment systems based on rule reasoning and knowledge base according to claim 6, characterized in that: Step S245 calculates the similarity of fault diagnosis information in TestFaiLRUes. The specific steps are as follows: S2451: Based on the LRU name in OrderLruMap, retrieve all fault diagnosis information corresponding to the LRU association dictionary from TestFaiLRUes, denoted as CurrentFaiLRUes; S2452: Loop through all trace records and record the information currently being traced as Trace; S2453: If Trace is a fault code, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and fault codes in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. S2454: If Trace is a function, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and function information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. S2455: If the Trace is a fault phenomenon, the cosine similarity algorithm is used to calculate the similarity between the Trace segmentation information and the fault phenomenon information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. S2456: If Trace is a component, then the cosine similarity algorithm is used to calculate the similarity between Trace segmentation information and component information in CurrentFaiLRUes, and the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result. S2457: If the Trace is a comprehensive rule, then the comprehensive rule is split, and the similarity is calculated and accumulated according to the rules in steps S2451 to S2456. Finally, the maximum similarity is combined with the fault diagnosis information to form a new record, which is added to the fault diagnosis result list result.

Images